KR102340902B1 - Apparatus and method for monitoring school zone - Google Patents

Abstract

Provided is an edge computing device capable of accurately and quickly detecting dangerous situations in a school zone, and the edge computing device includes: an input interface configured to receive an input image captured by at least one camera for capturing the school zone; a video buffer configured to store the input image; at least one first processor configured to input the input image captured by the at least one camera to a first machine learning model to acquire, from the first machine learning model, vehicle information and speed information detected from the input image and to detect an over-speed event in which the detected speed information exceeds a speed reference value; and a first communication unit configured to communicate with a first remote device installed in a school and a second remote device installed to be spaced apart from an installation point of the camera by a first reference distance in a direction opposite to a vehicle traveling direction, wherein the at least one first processor controls the first communication unit to transmit at least one of the vehicle information, the speed information, over-speed information on the over-speed event, or the input image or a combination thereof to the first and second remote devices.

Description

Apparatus and method for monitoring school zone}

Embodiments of the present disclosure relate to an apparatus and method for monitoring a school zone. Embodiments of the present disclosure disclose an edge computing device, a school zone monitoring method, and a remote device for monitoring the school zone.

Current traffic condition measurement-related technologies use a method of measuring section speed using a sensor installed on the road surface, a method of using a closed-circuit television (CCTV) image, and the like. However, when measuring the section speed using a sensor installed on the road surface, there are disadvantages in that the installation and management cost of the sensor is high, and the section speed is provided rather than the speed of the vehicle itself. The method using CCTV images is inconvenient in that a person has to directly view the image in the control system and determine whether or not traffic laws are violated. In addition, an unmanned enforcement system using drones is being implemented. In the case of a system using drones, drones were floated over highways to crack down on traffic violations using drones. In the drone-based method, a police officer directly judges whether a traffic law is violated by looking at the video filmed by the drone, and then contacts a nearby fellow police officer's patrol car to crack down on it. The method using drones is inconvenient because police officers have to read the video themselves.

Meanwhile, there is a recent demand for a transportation system for preventing accidents in school zones. The school zone has been designated as a child protection zone, and it is an area frequented by children, so it is important to provide drivers with information about the current situation in the school zone. In addition, a technology for monitoring a school zone in an institution such as a school is also required.

SUMMARY Embodiments of the present disclosure provide an edge computing device capable of accurately and quickly detecting a dangerous situation in a school zone, a school zone monitoring method, and a remote device.

According to an aspect of an embodiment of the present disclosure, an input interface for receiving an input image photographed from at least one camera for photographing a school zone; a video buffer for storing the input image; The input image captured by the at least one camera is input to a first machine learning model, and vehicle information and speed information detected from the input image are obtained from the first machine learning model, and the detected speed information is a speed reference. at least one first processor to detect a speeding event exceeding a value; and a first communication unit configured to communicate with a first remote device installed in a school, and a second remote device installed to be spaced apart from the installation point of the camera in a direction opposite to the vehicle traveling direction by a first reference distance, wherein the at least one second 1 The processor is configured to transmit, to the first remote device and the second remote device, at least one or a combination of the vehicle information, the speed information, the speeding event information for the speeding event, or the input image. 1 An edge computing device for controlling a communication unit is provided.

In addition, according to an embodiment of the present disclosure, the at least one first processor transmits the input image, the vehicle information, and the speed information to the first remote device, and to the second remote device, The first communication unit may be controlled to transmit the vehicle information, the speed information, and the speeding event information.

In addition, according to an embodiment of the present disclosure, the at least one first processor obtains, from the first machine learning model, information on the number of pedestrians in the school zone, and the first remote device and the second remote device It is possible to control the first communication unit to transmit the information on the number of pedestrians to at least one of them.

In addition, according to an embodiment of the present disclosure, the first communication unit communicates with a third remote device that provides pedestrian information, and the at least one first processor transmits the pedestrian information to the third remote device. can be transmitted

In addition, according to an embodiment of the present disclosure, the first communication unit includes a BCDMA communication module, and transmits the vehicle information, the speed information, and the speeding event information through the serial port of the BCDMA communication module, The input image may be transmitted through an Ethernet port.

Also, according to an embodiment of the present disclosure, the at least one first processor detects a license plate area and a driver area of the vehicle detected from the input image, and image processing for reducing visibility of the license plate area and the driver area to generate a corrected input image from which personal information is removed, and transmit the corrected input image through the first communication unit.

In addition, according to an embodiment of the present disclosure, the at least one first processor calculates the traffic amount information of the school zone based on the vehicle information and the speed information obtained from the first machine learning model, The traffic amount information may be transmitted to the first remote device or the second remote device through the first communication unit.

Also, according to an embodiment of the present disclosure, the at least one camera is installed for each lane of the school zone, and the at least one first processor includes an input image for each lane obtained from the at least one camera. Based on the , vehicle information and speed information for each lane may be acquired.

In addition, according to an embodiment of the present disclosure, when a speeding event is detected through the first communication unit, the at least one first processor transmits the speeding event information and the input image corresponding to the detected speeding event. transmit to a remote server, and receive speed information, vehicle information, and speeding event information detected for the input image from the remote server, and the remote server uses a second machine learning model to obtain the speed information from the input image and acquiring the vehicle information, and the first machine learning model may be a machine learning model to which a bypass path is applied to at least one layer of the second machine learning model.

According to another aspect of an embodiment of the present disclosure, receiving an input image photographed from at least one camera for photographing a school zone; storing the input image; inputting the input image captured by the at least one camera into a first machine learning model, and obtaining vehicle information and speed information detected from the input image from the first machine learning model; detecting a speeding event in which the detected speed information exceeds a speed reference value; and the at least one first processor is a first remote device installed in a school and a second remote device installed to be spaced apart from the installation point of the camera by a first reference distance in a direction opposite to the vehicle travel direction, the vehicle information, the Provided are speed information, speeding event information for the overspeeding event, or a method for controlling the Hue computing device.

According to another aspect of an embodiment of the present disclosure, receiving an input image photographed by at least one camera installed in the school zone; receiving vehicle information, speed information, and pedestrian information detected from the input image from a remote edge computing device; generating monitoring information for the school zone based on the input image, the vehicle information, the speed information, and the pedestrian information; transmitting the monitoring information to at least one client device; and transmitting at least one of the input image, the vehicle information, the speed information, the pedestrian information, and the monitoring information to a remote server.

Further, according to an embodiment of the present disclosure, the monitoring information includes school zone risk information, the pedestrian information includes information on the number of pedestrians, and the method includes: calculating the number of vehicles in the school zone; calculating school zone risk information based on whether the number of vehicles in the school zone exceeds a reference value for the number of vehicles and whether the number of pedestrians in the school zone exceeds a reference value for the number of pedestrians; and generating and outputting a traffic map guide based on the school zone risk information.

Also, according to an embodiment of the present disclosure, the traffic guidance guide may include a first-level guide that requests a teacher's guidance to and from school, and a second-level guide that requests a traffic map from a traffic police.

Further, according to an embodiment of the present disclosure, the method includes: collecting weather information; and collecting time information, wherein the calculating of the school zone risk information may further include calculating the school zone risk information based on the weather information and the time information.

Further, according to an embodiment of the present disclosure, the method may include: detecting the number of parked and stopped vehicles in an outermost lane; and calculating the school zone risk level information further based on whether the number of parked and stopped vehicles exceeds a parked and stopped reference value.

Also, according to an embodiment of the present disclosure, the edge computing device includes a first machine learning model for generating the vehicle information and the speed information from the input image, and the method includes: receiving speeding event information in which a vehicle having a speed exceeding a reference value is detected; transmitting at least one input image in which the speeding event information is detected to the remote server; receiving speed information, vehicle information, and speeding event information detected with respect to the input image from the remote server; and updating the monitoring information with the speed information, the vehicle information, and the speeding event information received from the remote server, wherein the remote server receives the information from the input image using a second machine learning model. Acquire speed information and the vehicle information, and the first machine learning model may be a machine learning model to which a bypass path is applied to at least one layer of the second machine learning model.

According to another aspect of an embodiment of the present disclosure, receiving an input image from at least one camera installed in a school zone, and receiving vehicle information, speed information, and pedestrian information detected from the input image from a remote edge computing device a communication unit for receiving; and generating monitoring information for the school zone based on the input image, the vehicle information, the speed information, and the pedestrian information, and transmitting the monitoring information to at least one client terminal through the communication unit, A school zone monitoring apparatus including a processor for transmitting at least one of the input image, the vehicle information, the speed information, the pedestrian information, and the monitoring information to a remote server through a communication unit is provided.

Also, according to an embodiment of the present disclosure, the edge computing device includes a first machine learning model that generates the vehicle information and the speed information from the input image, and the second processor is , Receives overspeed event information for detecting a vehicle having a speed exceeding a speed reference value, transmits at least one input image in which the speeding event information is detected, to the remote server, and from the remote server to the input image Receives the detected speed information, pedestrian information, and speeding event information, and updates the monitoring information with the speed information, the vehicle information, and the speeding event information received from the remote server, and the remote server includes a second Obtaining the speed information and the vehicle information from the input image by using a machine learning model, and the first machine learning model is a machine learning model in which a bypass path is applied to at least one layer of the second machine learning model can

According to embodiments of the present disclosure, there is an effect that can provide an edge computing device, a school zone monitoring method, and a remote device capable of accurately and quickly detecting a dangerous situation in a school zone.

1 is a view showing a school zone monitoring system according to an embodiment of the present disclosure.
2 is a diagram illustrating structures of an edge computing device, a first remote device, a second remote device, and a third remote device according to an embodiment of the present disclosure.
3 is a diagram showing the structure of a BCDMA module according to an embodiment of the present disclosure.
4 is a diagram illustrating a process of processing an input image in an edge computing device according to an embodiment of the present disclosure.
5 is a diagram illustrating an edge computing device, a camera, and a remote device according to an embodiment of the present disclosure.
6 is a diagram illustrating a control method of an edge computing device according to an embodiment of the present disclosure.
7 is a diagram illustrating a structure of an edge computing device according to an embodiment of the present disclosure.
8 is a diagram illustrating a structure of a first processor according to an embodiment of the present disclosure.
9 is a diagram illustrating an image processing process according to an embodiment of the present disclosure.
10 is a diagram illustrating a camera calibration process according to an embodiment of the present disclosure.
11 is a diagram illustrating a camera calibration process according to an embodiment of the present disclosure.
12 is a diagram for describing an object detection process according to an embodiment of the present disclosure.
13 is a diagram illustrating an object detection process 914 according to an embodiment of the present disclosure.
14 is a diagram for describing a tracklet information generation process according to an embodiment of the present disclosure.
15 is a diagram illustrating the structure of a tracklet network according to an embodiment of the present disclosure.
16 is a diagram illustrating a graph model according to an embodiment of the present disclosure.
17 is a diagram illustrating a process of calculating a relation between objects according to an embodiment of the present disclosure.
18 is a diagram illustrating a graph model generation process and a clustering process according to an embodiment of the present disclosure.
19 is a diagram illustrating a result image according to an embodiment of the present disclosure.
20 is a diagram showing the structure of a school zone monitoring system according to an embodiment of the present disclosure.
21 is a flowchart illustrating a school zone monitoring method according to an embodiment of the present disclosure.
22 is a diagram illustrating a process of generating school zone risk information and a traffic map guide according to an embodiment of the present disclosure.
23 is a view for explaining a process of generating school zone risk information and a traffic map guide from school zone environment information, according to an embodiment of the present disclosure.

This specification clarifies the scope of the claims of the present disclosure, and describes the principles of the embodiments of the present disclosure so that those of ordinary skill in the art to which the embodiments of the present disclosure pertain can practice the embodiments of the present disclosure and discloses embodiments. The disclosed embodiments may be implemented in various forms.

Like reference numerals refer to like elements throughout. This specification does not describe all elements of the embodiments, and general content in the technical field to which the embodiments of the present disclosure pertain or overlapping between the embodiments will be omitted. As used herein, the term 'part' (part, portion) may be implemented in software or hardware, and according to embodiments, a plurality of 'parts' may be implemented as one element (unit, element), or one 'part' It is also possible that ' includes a plurality of elements. Hereinafter, with reference to the accompanying drawings, embodiments of the present disclosure, and the principle of operation of the embodiments will be described.

1 is a view showing a school zone monitoring system according to an embodiment of the present disclosure.

A school zone is a children's protection area installed around a kindergarten or elementary school, and it means a commuter path within 300 meters from the school's main gate. Safety signs, road reflectors, and speed bumps are installed in school zones to protect children and reduce damage from traffic accidents. Also, vehicles cannot be parked or stopped within the school zone, and the speed is limited to 30 km/h or less.

In school zones, measures are taken for school zone safety by various entities such as schools, police, and drivers. For the safety of children in the school zone, the school operates the Green Mothers Association, and conducts guidance activities for child safety by faculty or teachers. For the safety of the school zone, the police dispatch traffic police to the school zone during commuting hours to organize traffic in the school zone and provide guidance for children's safety. Drivers must comply with traffic laws such as speed limits and parking bans applied to child protection zones, and pay attention to forward attention to protect children.

Embodiments of the present disclosure disclose a school zone monitoring system (100). The school zone monitoring system 100 includes an edge computing device 110 , a first remote device 120 , a remote server 130 , a second remote device 140 , and a third remote device 150 .

The edge computing device 110 , the second remote device 140 , and the third remote device 150 are disposed within the school zone. The edge computing device 110 includes a camera or is disposed adjacent to the camera. The second remote device 140 and the third remote device 150 are disposed to be spaced apart from the edge computing device 110 by a predetermined distance.

The edge computing device 110 corresponds to an electronic device that processes an input image. The edge computing device 110 may be disposed in a predetermined structure of the school zone.

The second remote device 140 is disposed to be spaced apart from the edge computing device 110 in a direction opposite to the vehicle traveling direction. The second remote device 140 displays information such as a speed limit, a detected speed of a vehicle, and a vehicle number. The second remote device 140 may be implemented in the form of a traffic sign, or may be implemented in the form of a circuit box embedded in the traffic sign. The second remote device 140 is disposed to be spaced apart from the edge computing device 110 in consideration of the safety distance of the vehicle, for example, may be disposed to be spaced apart from the edge computing device 110 by 20 m.

The third remote device 150 is disposed to be spaced apart from the edge computing device 110 in a direction opposite to the vehicle traveling direction. The third remote device 150 may be implemented in the form of a traffic sign board, or may be implemented in the form of a circuit box embedded in the traffic sign. The third remote device 140 displays the pedestrian information detected by the edge computing device 110 . The third remote device 140 is disposed to be spaced apart from the edge computing device 110 in consideration of the safety distance of the vehicle, for example, may be disposed to be spaced apart from the edge computing device 110 by 20 m.

A first remote device 120 is deployed at the school 20 and communicates with a client device 122 . The client device 122 is a device possessed by a school-related entity, such as a teacher, school staff, or sheriff, and may be implemented in the form of various electronic devices, such as a mobile phone, a tablet PC, a desktop PC, and a laptop PC. The first remote device 120 communicates with the edge computing device 110 and receives at least one or a combination of an input image, vehicle information, speed information, and lane information from the edge computing device 110 . The first remote device 120 generates monitoring information for the school zone and transmits the monitoring information to the client device 122 .

The remote server 130 is a server that collects school zone information and is connected to the first remote device 120 through the network 160 . The remote server 130 may correspond to a predetermined server device or a cloud server. The remote server 130 may receive at least one of an input image, speed information, vehicle information, and event information or a combination thereof from the first remote device 120 .

According to an embodiment, the remote server 130 may communicate with the edge computing device 110 . The remote server 130 may receive event information, an input image, speed information, and vehicle information from the edge computing device 110 , and may verify the event information. The configuration for verification of event information will be described in detail below.

2 is a diagram illustrating structures of an edge computing device, a first remote device, a second remote device, and a third remote device according to an embodiment of the present disclosure.

The edge computing device 110 includes a first processor 210 , an input interface 214 , a video buffer 216 , and a first communication unit 218 . The first processor 210 executes the first machine learning model 212 to perform the operation of the first machine learning model 212 .

The first processor 210 controls the overall operation of the edge computing device 110 . The first processor 210 may be implemented as one or more processors. The first processor 210 may execute an instruction or a command stored in the memory to perform a predetermined operation.

The first machine learning model 212 includes a deep neural network model. According to an embodiment, the deep neural network model may be operated in the first processor 210 by executing a computer program stored in the memory. For example, the deep neural network model may be implemented as a software module in the first processor 210 . According to another embodiment, the first processor 210 may include a separate dedicated processor for executing the deep neural network model.

According to another embodiment, the first processor 210 may use a deep neural network model operating in an external device such as a server. In this case, the first processor 210 may transmit an input image to be input to the deep neural network model to an external device and receive an output of the deep neural network model from the external device. To this end, the first processor 210 may communicate with an external device using the first communication unit 218 provided in the image edge computing device 110 .

The first machine learning model 212 may be learned using training data based on one or more nodes and an operation rule between nodes. A node structure, a layer structure, and an operation rule between nodes may be variously determined according to an embodiment. The first machine learning model 212 includes hardware resources such as one or more processors, memories, registers, summation processing units, or multiplication processing units, and operates the hardware resources based on a parameter set applied to each hardware resource. To this end, the first processor 210 operating the first machine learning model 212 may perform a task of allocating hardware resources or resource management processing for each operation of the first machine learning model 212 . The first machine learning model 212 may have, for example, a structure such as a recurrent neural network (RNN), a long short-term memory (LSTM), or the like.

The first processor 210 receives the input image, detects an object corresponding to the vehicle from at least some frames of the input image, and generates detected vehicle speed information. The first processor 210 executes the first machine learning model 212 to generate object identification information and object speed information. According to an embodiment, the first processor 210 may sample the frame of the input video at a second frame rate lower than the first frame rate of the input image, and input it to the first machine learning model. Also, the first processor 210 detects an event in which the speed of the object exceeds the first reference value, and generates overspeed event information.

The overspeed event information is information indicating whether an object having a speed exceeding a first reference value or a second reference value is detected. The speeding event information may include information on an object, information on a frame in which a speeding event occurs, and detected speed information. The speeding event information may include event detection information indicating that a speeding event is detected and event non-detection information indicating that a speeding event is not detected. The speeding event information may be generated for each frame. The information about the object may include, for example, information about coordinates, a horizontal length, and a vertical length of a region corresponding to the object in the frame. The information on the frame in which the speeding event occurs may include information such as a frame number or a time corresponding to the frame.

According to an embodiment, the first processor 210 generates vehicle information from an input image. The first processor 210 obtains vehicle information from the input image using an object recognition algorithm or a text recognition algorithm. The vehicle information includes, for example, at least one of a vehicle model, a license plate, and a vehicle model, or a combination thereof.

According to an embodiment, the first processor 210 generates pedestrian information from the input image. The first processor 210 may obtain pedestrian information by inputting the input image to the first machine learning model 212 . The pedestrian information represents area information corresponding to the pedestrian detected in the input image. The first processor 210 may generate additional information such as the number of pedestrians and the density of pedestrians based on the pedestrian information output from the first machine learning model 212 .

The input interface 214 receives an input image photographed from at least one camera for photographing the school zone. According to an embodiment, the edge computing device 110 may include a camera in the edge computing device 110 . In this case, the input interface 214 receives an input image from a camera built into the edge computing device 110 . According to another embodiment, it may be connected to a camera disposed outside the edge computing device 110 to receive an input image through the input interface 214 . In this case,

The camera captures the school zone and transmits the captured image data to the edge computing device 110 . The camera is placed with a Field of View (FOV) set to take pictures of the school zone. According to an embodiment, the camera may correspond to an existing CCTV camera.

The school zone monitoring system 100 may include one or a plurality of cameras. According to one embodiment, the school zone monitoring system 100 is a first camera for photographing a vehicle driving in the roadway of the school zone, and a second camera for photographing the front of the vehicle driving toward the crosswalk around the crosswalk. may include The FOV of the second camera may be set to photograph a vehicle driving toward the crosswalk, including the crosswalk. As another example, the FOV of the second camera may be set to photograph up to a stop line in front of the crosswalk without including the crosswalk.

The input interface 214 may correspond to an input device or a communication unit having a predetermined standard for receiving image data from a camera capable of receiving image data from the outside. The input interface 214 transmits the input image data to the first processor 210 and the video buffer 216 . The input image data corresponds to the input image.

The video buffer 216 stores an input image. The video buffer 116 may include a storage medium capable of storing a moving picture. The video buffer 216 is, for example, a flash memory type, a hard disk type, a multimedia card micro type, or a card type memory (eg SD or XD). memory, etc.), RAM (Random Access Memory), SRAM (Static Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory), It may include at least one type of storage medium among a magnetic memory, a magnetic disk, and an optical disk.

The first communication unit 218 may communicate with an external device by wire or wirelessly. The first communication unit 218 communicates with the first remote device 120 , the second remote device 140 , and the third remote device 150 . Also, according to an embodiment, the first communication unit 218 may communicate with a remote server.

The first communication unit 218 communicates with an external device by wire or wirelessly. According to an embodiment, the first communication unit 218 may perform short-range communication, for example, BCDMA, LoRa (Long Range), Wi-Sun (Wi-SUN), Bluetooth, BLE (Bluetooth Low Energy), Near Field Communication, WLAN (Wi-Fi), Zigbee, Infrared Data Association (IrDA) communication, WFD (Wi-Fi Direct), UWB (ultra wideband), Ant+ communication, etc. are available. . As another example, the first communication unit 218 may perform wide area communication such as a mobile communication network, optical communication, or Ethernet. The first communication unit 218 may include at least one of at least one short-range communication module and at least one wide area communication module, or a combination thereof.

According to an embodiment, the first communication unit 218 may include a binary code division multiple access (BCDMA) module. The first communication unit 218 may include a master BCDMA module, and the first remote device 120 , the second remote device 140 , and the third remote device 150 may include a slave BCDMA module. The first communication unit 218 and the first remote device 120 , the first communication unit 218 and the second remote device 140 , and the first communication unit 218 and the third remote device 150 are connected via a BCDMA module. can communicate

According to an embodiment, the first processor 210 transmits an input image, vehicle information, and speed information to the first remote device 120 , and transmits vehicle information, speed information, and/or information to the second remote device 140 , to the second remote device 140 . and controls the first communication unit 218 to transmit speeding event information.

The first remote device 120 includes a second processor 220 , a second communication unit 222 , and a network video recorder (NVR) 224 .

The second processor 220 controls the overall operation of the first remote device 120 . The second processor 220 may be implemented as one or more processors. The second processor 220 may execute an instruction or a command stored in the memory to perform a predetermined operation.

According to an embodiment, the second processor 220 may monitor information for the school zone based on at least one of an input image received from the edge computing device 110 , vehicle information, speed information, and pedestrian information, or a combination thereof. to create The second processor 220 receives at least one or a combination of an input image, vehicle information, speed information, and pedestrian information from the edge computing device 110 through the second communication unit 222 , and based on the received information to create monitoring information. Monitoring information is information indicating the current situation and risk of the school zone. The monitoring information may include, for example, vehicle congestion, pedestrian information, the number of parked and stopped vehicles, speeding vehicle information, and the like. In addition, the monitoring information may include school zone risk information. The school zone risk information is information indicating the risk of the school zone, and is determined based on at least one of the number of vehicles, the number of pedestrians, weather, time information, and the number of parked and stopped vehicles.

Also, according to an embodiment, the second processor 220 may generate and output a traffic map guide based on the school zone risk information. The traffic map guide requests traffic guidance from the school's supervising teacher or traffic police based on the school zone risk information. The traffic guidance guide includes, for example, a first-level guide that requests guidance from a tutor to provide guidance to and from school, and a second-level guide that requests a traffic guidance from a traffic police. The second processor 220 outputs the first level guide or the second level guide to the client device 122 based on the school zone risk information. According to an embodiment, the second processor 220 may determine to which client device 122 to transmit the traffic map guide based on the school zone risk information. For example, when the first level guide is generated, the traffic map guide is transmitted to the teacher's client device 122, and when the second level guide is generated, the traffic map guide is transmitted to the police's client device 122 can be transmitted

In addition, according to an embodiment, the first remote device 120 includes an output interface such as a display unit, and outputs school zone risk information or a traffic map guide through the output interface.

The second communication unit 222 communicates with an external device by wire or wirelessly. According to one embodiment, the second communication unit 222 may perform short-range communication, for example, BCDMA, LoRa (Long Range), Wi-Sun (Wi-SUN), Bluetooth, BLE (Bluetooth Low Energy), Near Field Communication, WLAN (Wi-Fi), Zigbee, Infrared Data Association (IrDA) communication, WFD (Wi-Fi Direct), UWB (ultra wideband), Ant+ communication, etc. are available. . As another example, the second communication unit 222 may perform wide area communication such as a mobile communication network, optical communication, or Ethernet. The second communication unit 222 may include at least one of at least one short-range communication module and at least one wide area communication module, or a combination thereof.

According to an embodiment, the second communication unit 222 may include a BCDMA module. As described above, the BCDMA module of the second communication unit 222 may correspond to the slave BCDMA module of the BCDMA module of the first communication unit 218 .

The NVR 224 records the input image received from the edge computing device 110 . The edge computing device 110 receives a video stream corresponding to the input image through the second communication unit 222 , and stores the video stream in the NVR 224 . The second processor 220 may transmit the input image stored in the NVR 224 to the remote server 130 .

The second remote device 140 includes a third communication unit 230 and a display 232 .

The third communication unit 230 communicates with an external device by wire or wirelessly. According to an embodiment, the third communication unit 230 may perform short-range communication, for example, BCDMA, LoRa (Long Range), Wi-Sun (Wi-SUN), Bluetooth, BLE (Bluetooth Low Energy), Near Field Communication, WLAN (Wi-Fi), Zigbee, Infrared Data Association (IrDA) communication, WFD (Wi-Fi Direct), UWB (ultra wideband), Ant+ communication, etc. are available. . As another example, the third communication unit 230 may perform wide area communication such as a mobile communication network, optical communication, or Ethernet. The third communication unit 230 may include at least one of at least one short-range communication module and at least one wide area communication module, or a combination thereof.

According to an embodiment, the third communication unit 230 may include a BCDMA module. As described above, the BCDMA module of the third communication unit 230 may correspond to the slave BCDMA module of the BCDMA module of the first communication unit 218 .

The display 232 may correspond to various types of display devices. According to an embodiment, the display 232 may include a light emitting diode (LED) display, a liquid crystal display, or the like. According to one embodiment, display 232 comprises a 7 segment display.

According to an embodiment, the second remote device 140 may further include a processor for controlling the operation of the second remote device 140 , and a memory for storing received data and predetermined information.

The second remote device 140 receives at least one or a combination of an input image, vehicle information, speed information, speeding event information, and pedestrian information from the edge computing device 110 . The second remote device 140 displays at least one of the received vehicle information, speed information, speeding event information, and pedestrian information on a predetermined area. For example, the second remote device 140 may simultaneously display the received vehicle information, speed information corresponding to the corresponding vehicle information, and speeding event information corresponding to the corresponding vehicle information on the display 232 . Also, the second remote device 140 may display pedestrian information on the display 232 . For example, the second remote device 140 may display information on the number of pedestrians, information on pedestrian attention indicating that there are many pedestrians, and the like.

According to an embodiment, the display 232 of the second remote device 140 includes a Variable Message Sign (VMS) and displays the pedestrian information received from the edge computing device 110 through the VMS. The second remote device 140 may display information on the number of pedestrians and information on pedestrian attention indicating that there are many pedestrians through the VMS.

According to an embodiment, the second remote device 140 displays an input image corresponding to the vehicle in which the speeding event is detected. In this case, the second remote device 140 may display the input image of the vehicle in which the speeding event is detected, vehicle information, and speed information together.

The third remote device 150 includes a fourth communication unit 240 and a pedestrian advance warning light 242 .

The fourth communication unit 240 communicates with an external device by wire or wirelessly. According to an embodiment, the fourth communication unit 240 may perform short-range communication, for example, BCDMA, LoRa (Long Range), Wi-Sun (Wi-SUN), Bluetooth, BLE (Bluetooth Low Energy), Near Field Communication, WLAN (Wi-Fi), Zigbee, Infrared Data Association (IrDA) communication, WFD (Wi-Fi Direct), UWB (ultra wideband), Ant+ communication, etc. are available. . As another example, the third communication unit 230 may perform wide area communication such as a mobile communication network, optical communication, or Ethernet. The third communication unit 230 may include at least one of at least one short-range communication module and at least one wide area communication module, or a combination thereof.

According to an embodiment, the fourth communication unit 240 may include a BCDMA module. As described above, the BCDMA module of the fourth communication unit 240 may correspond to the slave BCDMA module of the BCDMA module of the first communication unit 218 .

The pedestrian advance warning light 242 is a warning light indicating pedestrian information. The pedestrian advance warning light 242 may correspond to a blinking lamp or an LED display that blinks under a predetermined condition. The pedestrian advance warning light 242 may display information such as presence of pedestrians, the number of pedestrians, or the density of pedestrians.

According to an embodiment, the third remote device 150 may further include a processor for controlling the operation of the third remote device 150 and a memory for storing received data and predetermined information.

The third remote device 150 receives pedestrian information from the edge computing device 110 . Pedestrian information includes information such as presence of pedestrians, the number of pedestrians, or pedestrian density. The third remote device 150 displays the received pedestrian information in a predetermined area. For example, the third remote device 150 may display the received pedestrian information on the pedestrian advance warning light 242 .

According to an embodiment, the edge computing device 110 detects, from the input image, the number of people on the driving road of the school zone, that is, on the road, and outputs the detected number of people in the road as pedestrian information. The number of people present on the roadway may include the number of people in the crosswalk and the number of people outside the crossing. The edge computing device 110 may output pedestrian information including the number of people in the road to the third remote device 150 and output the number of people in the road as the number of pedestrians.

3 is a diagram showing the structure of a BCDMA module according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the first communication unit 218 , the second communication unit 222 , the third communication unit 230 , and the fourth communication unit 240 may include a BCDMA module. Here, the first communication unit 218 may correspond to the master BCDMA module 310 , and the second communication unit 222 , the third communication unit 230 , and the fourth communication unit 240 may correspond to the slave BCDMA module 320 . have. The edge computing device 110 has master authority over the first remote device 120 , the second remote device 140 , and the third remote device 150 , and the first remote device 120 , the second remote device 140 and the third remote device 150 may correspond to a slave device of the edge computing device 110 . The edge computing device 110 transmits a control signal to each of the first remote device 120 , the second remote device 140 , and the third remote device 150 through the master BCDMA module 310 , and the first remote device Each of the device 120 , the second remote device 140 , and the third remote device 150 may be controlled according to a control signal received from the edge computing device 110 . With this configuration, the processing burden and required resources of the first remote device 120 , the second remote device 140 , and the third remote device 150 are remarkably reduced, and the school zone monitoring system 100 is constructed. The cost can be reduced, and the burden of increasing the number of remote devices can be reduced.

The edge computing device 110 generates control signals for each of the first remote device 120 , the second remote device 140 , and the third remote device 150 , and the first remote device 120 , the second Control signals for each of the remote device 140 and the third remote device 150 are transmitted to each remote device via the master BCDMA module 310 .

The edge computing device 110 transmits an input image, vehicle information, speed information, and pedestrian information to the first remote device 120 , and the input image from the first remote device 120 is recorded in the NVR 224 , and input A control signal for controlling the first remote device 120 may be transmitted so that vehicle information and speed information are displayed on the image. The first remote device 120 operates based on a control signal received from the edge computing device 110 in at least some operations.

The edge computing device 110 transmits vehicle information, speed information, speeding event information, and pedestrian information to the second remote device 140 , and the second remote device 140 uses a predefined method for vehicle information and speed information. , a control signal for controlling the second remote device 140 to display the speeding event information, and the pedestrian information. The second remote device 140 operates based on a control signal received from the edge computing device 110 in at least some operations.

The edge computing device 110 transmits pedestrian information to the third remote device 150 , and controls the third remote device 150 to display the pedestrian information in a predefined manner in the third remote device 150 . signal can be transmitted. The third remote device 150 operates based on a control signal received from the edge computing device 110 in at least some operations.

According to an embodiment, the master BCDMA module 310 includes a first serial port 312 and a first Ethernet port 314 . The slave BCDMA module 320 includes a second serial port 322 and a second Ethernet port 324 . The master BCDMA module 310 transmits vehicle information, speed information, pedestrian information, or speeding event information through the first serial porter 312 and transmits an input image through the first Ethernet porter 314 . The serial port of the BCDMA module is a port used to transmit information for checking the setting and operation status log. According to an embodiment of the present disclosure, vehicle information, speed information, pedestrian information, or speeding information, which is not large in data size, Event information is transmitted to the second serial port 322 using the first serial port 312, and an input image having a large data size is transmitted from the first Ethernet port 314 to the second Ethernet port 324. , which has the effect of simplifying data processing. In the case of an input image, since the input of the NVR 224 uses the second Ethernet port 324 , it is possible to directly connect the existing NVR 224 and the edge computer device 110 without any other data processing. In addition, in the case of detection data such as vehicle information, speed information, pedestrian information, or speeding event information, by transmitting the data through the first serial port 312 and the second serial port 324, the receiving side receives the serial data. There is an effect that data can be acquired through data processing such as parsing using a low-cost micro control unit (MCU). For example, in the case of the second remote device 140 , the MCU can directly control the 7-segment display to display data on the display, thereby remarkably reducing required hardware resources.

4 is a diagram illustrating a process of processing an input image in an edge computing device according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the edge computing device 110 detects the license plate area 420 and the driver area 430 of the vehicle 410 detected from the input image. The first processor 210 of the edge computing device 110 detects the vehicle 410 from the input image. The first processor 210 detects a region corresponding to the front of the vehicle from the input image. The first processor 210 detects the license plate area 320 and the driver area 430 from the image data corresponding to the front of the vehicle 410 . The first processor 210 detects the rectangular license plate area 420 of the central lower area of the vehicle area based on the information on the structure of the vehicle, and the driver area 430 of the right area of the vehicle windshield above the vehicle area. ) is detected.

The first processor 210 performs image processing for reducing visibility on the license plate area 420 and the driver area 430 detected from the input image. For example, the first processor 210 reduces visibility by modifying the data values of the license plate area 420 and the driver area 430 to predetermined default values or performing mosaic processing, blurring processing, and the like. According to an embodiment, the first processor 210 reduces the visibility of the driver area 430 , reduces the visibility of the license plate area 420 , or reduces the visibility of the driver area 430 and the license plate area 420 . can be reduced together.

The first processor 210 transmits an input image with reduced visibility of at least one of the license plate area 420 and the driver area 430 to the first remote device 120 or the second remote device 140 .

5 is a diagram illustrating an edge computing device, a camera, and a remote device according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the school zone monitoring system 100 installs a camera in each lane of the school zone, and vehicle information corresponding to each lane, speed information, and overspeed in a remote device corresponding to each lane Output event information. The school zone monitoring system 100 includes a first lane camera 510 , a second lane camera 520 , and a third lane camera 530 corresponding to each lane. Each lane camera may be provided to correspond to the number of lanes, and an FOV may be set and installed to photograph the front of the driving vehicle in each lane. According to an embodiment, the camera is arranged such that one camera captures a plurality of lanes (eg, two lanes, three lanes, etc.) according to the FOV of the camera, and the edge computing device 110 is configured in each camera. By defining each lane area, a vehicle may be detected from each lane.

The remote devices are provided and arranged to correspond to each lane, and include, for example, a first lane remote device 520 , a second lane remote device 522 , and a third lane remote device 524 . Here, the first lane remote device 520 , the second lane remote device 522 , and the third lane remote device 524 disposed in each lane correspond to the second remote device 140 of FIG. 2 . Accordingly, the first lane remote device 520 , the second lane remote device 522 , and the third lane remote device 524 each include a third communication unit 230 and a display 232 . The first lane remote device 520, the second lane remote device 522, and the third lane remote device 524 are disposed on top of each lane and are disposed to face the front of a vehicle traveling in each lane, Speed information, speeding event information, and the like may be provided to vehicles traveling in each lane.

The edge computing device 110 receives input images from the first lane camera 510 , the second lane camera 512 , and the third lane camera 514 , vehicle information for each lane, speed information, and overspeed Detect event information. The edge computing device 110 generates vehicle information, speed information, and speeding event information for each lane, respectively, and transmits the generated information to a remote device corresponding to each lane.

According to an embodiment, the edge computing device 110 defines lane information corresponding to the input image from identification information of a camera that has transmitted the input image, and vehicle information, speed information, and speeding event from the input image of each lane. create information The edge computing device 110 transmits the vehicle information, the speed information, and the speeding event information to the remote devices 520 , 522 , and 524 of the lane corresponding to the corresponding vehicle information, the speed information, and the speeding event information.

According to another embodiment, at least one camera covering a plurality of lanes is disposed, the edge computing device 110 detects a lane from an input image, detects a vehicle from each lane, and information on the detected vehicle lane; Generate vehicle information, speed information, and speeding event information. The edge computing device 110 transmits vehicle information, speed information, and speeding event information to the remote devices 520 , 522 , and 524 corresponding to each lane.

6 is a diagram illustrating a control method of an edge computing device according to an embodiment of the present disclosure.

Each step of the edge computing device control method of the present disclosure may be performed by various types of electronic devices having a processor and a communication unit and using a machine learning model. The present disclosure will focus on an embodiment in which the edge computing device 110 according to embodiments of the present disclosure performs an edge computing device control method. Accordingly, the embodiments described for the edge computing device 110 are applicable to embodiments for the edge computing device control method, and on the contrary, the embodiments described for the edge computing device control method are for the edge computing device 110 . Applicable to the embodiments. The edge computing device control method according to the disclosed embodiments is not limited to being performed by the edge computing device 110 disclosed in the present disclosure, and may be performed by various types of electronic devices.

The edge computing device 110 receives an input image photographed from at least one camera for photographing the school zone (S602). The camera may be embedded in the edge computing device 110 or disposed adjacent to the edge computing device 110 . The edge computing device 110 receives an input image from a camera through a predetermined input interface.

Next, the edge computing device 110 stores the received input image. The edge computing device 110 may include a predetermined video buffer and store the received input image in the video buffer.

Next, the edge computing device 110 inputs the input image to the first machine learning model, obtains vehicle information and speed information detected from the input image from the first machine learning model, and the detected speed information is a speed reference value Detecting a speeding event exceeding , generates speeding event information (S606).

The first machine learning model is a machine learning model trained to obtain vehicle information and speed information from an input image. The first machine learning model is executed by the first processor of the edge computing device 110 or is provided in the cloud server and communicates with the edge computing device 110 to obtain vehicle information and speed information from the input image. can be output as

The edge computing device 110 generates speeding event information based on vehicle information and speed information obtained from the first machine learning model. When the speed value of speed information obtained from the first machine learning model exceeds a predetermined speed reference value, the edge computing device 110 determines that a speeding event has occurred and generates overspeed event information.

Also, according to an embodiment, the edge computing device 110 generates pedestrian information from the input image. The edge computing device 110 generates pedestrian information from the input image. The edge computing device 110 may obtain pedestrian information by inputting the input image to the first machine learning model. The pedestrian information represents area information corresponding to the pedestrian detected in the input image. The edge computing device 110 may generate additional information such as the number of pedestrians and the density of pedestrians based on the pedestrian information output from the first machine learning model.

Next, the edge computing device 110 receives at least one or a combination of an input image, vehicle information, speed information, speeding event information, and pedestrian information to the first remote device 120 and the second remote device 140 . It is transmitted (S608).

In addition, the embodiments of the present disclosure described with reference to FIGS. 1 to 23 may be applied to an edge computing device control method. Also, operations of the edge computing device 110 described with reference to FIGS. 1 to 23 may be added as steps of the edge computing device control method.

7 is a diagram illustrating a structure of an edge computing device according to an embodiment of the present disclosure. In FIG. 7 , the camera 710 , the first processor 210a , and the video buffer 216 of the edge computing device 110a will be mainly described, and other configurations will be omitted.

The blocks defined in the first processor 210a in the present disclosure are only examples of hardware blocks or software processing units for performing the embodiments of the present disclosure, and embodiments of the present disclosure are performed in various ways other than the processing units disclosed in the present disclosure. A processing unit may be defined.

The first processor 210a of the edge computing device 110a may include a codec 720 , a sampler 730 , a first machine learning model 226 , and a first speed analyzer 740 . Each block of the first processor 210a may correspond to a hardware block or a software block. For example, the first processor 210a may include a dedicated processor corresponding to the codec 720 or a dedicated processor corresponding to the first machine learning model 226 .

The codec 720 decodes an input video input from the camera 710 into a predetermined standard. The codec 720 may support standards such as, for example, AVI, Moving Picture Experts Group (MPEG), MOV, and Window Media Video (WMV). The codec 720 generates a plurality of input frames by decoding the input video.

The sampler 730 receives a plurality of input frames from the codec 720 and performs sampling processing on the plurality of input frames. For example, the sampler 730 may receive an input frame at 30 fps and sample at 3 fps. A sampling rate of the sampler 730 may vary according to an embodiment.

The first machine learning model 226 generates and outputs identification information and speed information of an object in each input frame from a plurality of input frames input from the sampler 730 . The plurality of input frames may be converted into input vectors required by the first machine learning model 226 through predetermined pre-processing, and the input vectors may be input to the first machine learning model 226 . For example, in the pre-processing, at least one object corresponding to a vehicle may be detected from an input frame, and information on the at least one object may be input to the first machine learning model 226 . The identification information and speed information of the object generated by the first machine learning model 226 are inserted into the input frame and output. For example, in the image data of the input frame, a box indicating an area of an object, information indicating identification information of an object (eg, a color of a box), and a speed of each object may be displayed together. The first processor 210a may insert object identification information and speed information into the image data through post-processing of the output of the first machine learning model 226 .

The first speed analyzer 740 receives the identification information and speed information of the object generated from the first machine learning model 226, determines whether the speed of each object exceeds a speed reference value, and information about the speeding event to create The speed reference value may be determined based on a prescribed speed (eg, 30 km/h) of the child protection zone. For example, the speed reference value may be determined as a value corresponding to 110% of the prescribed speed of the child protection zone. The first speed analyzer 740 detects an event in which the speed of each object in the input frame exceeds a speed reference value. The first speed analyzer 740 may generate event detection information indicating that a speeding event is detected or event non-detection information indicating that an event is not detected according to the processing result. The event detection information may include identification information and frame information (or time information) of an object in which a speeding event is detected.

The video buffer 116 of the edge computing device 110a stores an input image including a plurality of input frames decoded and generated by the codec 720 . The video buffer 116 outputs the input image to at least one of the first remote device 120 , the second remote device 140 , and the remote server 130 , or a combination thereof.

8 is a diagram illustrating a structure of a first processor according to an embodiment of the present disclosure.

The first processor 210b according to an embodiment of the present disclosure includes a sampler 730 , a preprocessor 810 , a first machine learning model 212 , a postprocessor 820 , and a first speed analyzer 740 . may include Since the sampler 730 , the first machine learning model 212 , and the first speed analyzer 740 of FIG. 8 are the same as those described in FIG. 7 , in FIG. 8 , the preprocessor 810 and the postprocessor 820 . will be explained based on

The preprocessor 810 receives the input frame output from the sampler 730 , detects an object, and generates an input vector including object information. The preprocessor 810 may generate an object corresponding to the vehicle from the input frame. The preprocessor 810 may extract location information of an object, information about width and height of an object region, and appearance information from an input frame, and generate an input vector including location information, width, height, and appearance information. When a plurality of objects are detected from the input frame, location information, width, height, and appearance information are generated for each object.

According to an embodiment, the preprocessor 810 may be implemented as a machine learning model. According to another embodiment, the preprocessor 810 may be implemented as a software module that performs a predetermined object detection algorithm.

The first machine learning model 212 receives the input vector and the input frame generated by the preprocessor 810, and generates object identification information and object speed information.

The post-processing unit 820 generates a plurality of connecting lines accumulating image data of an object region corresponding to each object based on the object identification information and the object velocity information output from the first machine learning model 212, Generates an output data structure including relevance information indicating path similarity between a plurality of connecting lines. Each node corresponds to each object, and image data generated by cutting image data of an object region may be accumulated in each node. The output data structure may correspond to a model in which each node is defined and a property of a connection line between each node is changed according to relation information between each node.

The post-processing unit 820 may additionally generate path information of each object based on the graph model. The path information of each object is information showing a trajectory of each object moving for a predetermined time in the past. The post-processing unit 820 generates relevance information based on the similarity of the path information.

The first speed analyzer 740 detects an event based on the object identification information and the object speed information generated by the first machine learning model 212 . The first speed analyzer 740 may detect an event in which the speed of the object exceeds the speed reference value based on the graph model generated by the post-processing unit 820 . The first speed analyzer 740 may detect an event based on speed information of each object accumulated in each node of the graph model.

9 is a diagram illustrating an image processing process according to an embodiment of the present disclosure.

The first processor 210b processes the input frame through pre-processing, machine learning model processing, and post-processing. In FIG. 9 , the pre-processing 910 , the machine learning model processing 920 , and the post-processing 930 performed by the first processor 210b will be described.

According to an embodiment of the present disclosure, the pre-processing 910 includes a camera calibration 912 , an object detection 914 , and an embedding process 916 . Machine learning model processing 920 includes processing by the first machine learning model. Post-processing 930 includes graph model generation processing 932 and clustering processing 934 .

10 is a diagram illustrating a camera calibration process according to an embodiment of the present disclosure.

The camera calibration process 912 is a process for converting the actual distance on the road into a rectangular form with respect to the input frame. The camera calibration process 912 may generate reference lines 1012a and 1012c indicating the same distance on a road and a reference line 1012b indicating a lane traveling direction on the road in the input frame 1010 . According to an embodiment, the reference lines 1012a , 1012b , and 1012c may be determined based on lane information detected in the input frame 1010 .

Next, the camera calibration process 912 may generate distance reference points 1022 indicating the distance on the road based on the lane travel direction. Based on the adjustment input frame 1020 on which the camera calibration process 912 has been performed, there is an effect of accurately measuring the inter-frame movement distance of the object from the input frame.

11 is a diagram illustrating a camera calibration process according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the camera calibration process 912 may perform a camera calibration process based on an image type of an input video or a road marking line. To this end, the camera calibration process 912 first identifies an image type or a road marking line based on the input image 1110 (operation 1112 ). The image type may include information on a type of camera equipment and a photographing composition. Road marking lines may include lanes, crosswalks, or stop lines. Next, the camera calibration process 912 performs camera calibration based on the image type or road marking line ( 1114 ).

The input image 1110 may be acquired by being photographed by various camera equipment. In addition, when capturing the input image, the composition of the camera equipment may be variously determined. The camera calibration process 912 performs the camera calibration process 912 based on image type information indicating information on at least one of a type of camera equipment or a shooting composition for capturing an input image, or a combination thereof.

The type of camera equipment may include, for example, a fixed CCTV camera, a rotational CCTV camera, or a drone camera. The photographing composition may include an oblique composition, a front composition, or an overhead photographing composition. The camera calibration process 912 may vary depending on whether the camera equipment is a stationary type, a rotation type, or a mobile type (eg, a drone). When the camera equipment is a fixed type, additional camera calibration processing 912 is not performed after the initial camera calibration processing 912, or the camera calibration processing 912 is performed at a predetermined period (eg, 1 hour). can When the camera equipment is a rotation type, when the camera equipment is rotated (eg, an operation such as pan or tilt) occurs, a camera calibration process 912 may be performed. In the case of rotating camera equipment, in addition to the case in which rotation occurs, the camera calibration process 912 may be performed at a predetermined cycle after the last camera calibration process 912 . In the case of mobile camera equipment such as a drone camera, the camera calibration process 912 may be performed in real time, for example, the camera calibration process 912 may be performed for every input frame after the sampling process.

In addition, the camera calibration process 912 varies depending on the shooting composition. The camera calibration process 912 defines the shape of a quadrangle generated by the reference lines 1012a, 1012b, and 1012c of the camera calibration process 912 differently depending on whether it is an oblique composition, a front composition, or an overhead composition.

In addition, the camera calibration process 912 may be determined based on the directions of the reference lines 1012a , 1012b , and 1012c based on road marking lines such as lanes and the arrangement of the distance reference points 1022 .

Referring again to FIG. 9 , preprocessing 910 includes object detection processing 914 . The object detection processing 914 detects an object corresponding to the vehicle from the input frame.

12 is a diagram for describing an object detection process according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the object detection process 914 detects at least one object 1222 corresponding to the vehicle from the input frame 1210 . The object detection processing 914 defines a block 1220 corresponding to an object region for an object detected in the input frame 1210, coordinates (x, y) of a reference point 1224 of the block, and a block 1220 ) defines the width (w) and the height (h) of the block. The reference point 1224 may be defined, for example, by the coordinates of the upper left corner of the block 1220 . Coordinates (x, y) indicate coordinates based on a horizontal axis and a vertical axis within the input frame 1210 .

Also, the object detection processing 914 may generate appearance information from the detected object. The appearance information may include, for example, information such as a vehicle type (eg, a passenger car, a bus, a truck, a sport utility vehicle (SUV), etc.) of the object, and a color.

13 is a diagram illustrating an object detection process 914 according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the object detection processing 914 performs an inter-frame overlap region processing 1330 and an appearance similarity calculation processing 1340 of the detected object from a plurality of sequentially inputted input frames.

The inter-frame overlap region processing 1330 detects an overlap region between temporally adjacent frames, and uses a processing result of a previous frame for the overlap region. The inter-frame overlap region processing 1330 detects an overlap region 1336 overlapping the first frame 1332 and the second frame 1334 that is a frame following the first frame 1332 , and the overlap region 1336 . , object detection processing is performed on the second frame 1334 based on the object detection processing result of the first frame 1332 . The overlapping region 1336 may be defined based on the detected object or may be defined based on image similarity.

The appearance similarity calculation processing 1340 detects objects having similar appearance information from a plurality of input frames. For example, the appearance similarity calculation processing 1340 calculates the appearance similarity between the object detected in the first frame and the object detected in the second frame. According to an embodiment, the appearance similarity may be generated based on a feature map generated using a CNN from an input frame. According to another embodiment, the appearance similarity may be calculated based on appearance information of the object of the first frame and appearance information of the object of the second frame.

Based on the object detection processing result generated based on the inter-frame overlap region processing 1330 and the appearance similarity calculation processing 1340 , the tracklet information V is generated.

Referring back to FIG. 9 , the preprocessing 910 performs a tracklet information generation process 916 based on the result of the object detection process 914 and the input frame on which the camera calibration process 912 is performed. The tracklet information generation processing 916 generates tracklet information 918 corresponding to an input vector input to the first machine learning model from the result of the object detection processing 914 and the input frame.

14 is a diagram for describing a tracklet information generation process according to an embodiment of the present disclosure.

The tracklet information generation process 916 receives an object detection result and an input frame, and performs a feature embedding process 1410 . The feature embedding process 1410 generates tracklet information 1418 from the object detection result and the input frame. The tracklet information 1418 is information corresponding to an input vector input to the tracklet network 1430 corresponding to the first machine learning model. The tracklet includes area information 1424 and appearance information 1426 about the area of the detected object in a format required by the tracklet network 1430 . The area information 1424 includes coordinates (x, y), a width (w), and a height (h) of a block corresponding to the object area. Appearance information 1426 includes appearance information previously generated in object detection processing 1414 . The tracklet information 1418 may include information 1422 about each object in one frame. The tracklet information 1418 may include information on a plurality of objects in a predetermined format.

Referring back to FIG. 9 , the tracklet information 918 generated by the tracklet information generation processing 916 is input to the machine learning model processing 920 together with the input frame. The machine learning model processing 920 processes the tracklet information and the input frame by the first machine learning model described above.

15 is a diagram illustrating the structure of a tracklet network according to an embodiment of the present disclosure.

The tracklet network 1500 has a DNN structure including a plurality of layers. The tracklet network 1500 may include a combination of a CNN structure and an RNN structure. The tracklet network 1500 includes an input layer 1510 , a plurality of hidden layers 1520 , and an output layer 1530 .

The input layer 1510 includes at least one layer 1514 . The input layer 1510 receives an input vector 1512 and an input frame, and generates at least one input feature map 1522 .

At least one feature map 1522 is input to the hidden layer 1520 and processed. The hidden layer 1520 is previously learned and generated by a predetermined machine learning algorithm. The hidden layer 1520 receives at least one input feature map 1522 and performs predetermined activation processing, pooling processing, linear processing, convolution processing, etc. to generate at least one output feature map 1526 . According to an embodiment, the hidden layer 1520 may include a processing layer corresponding to each input feature map 1522 . The feature map 1526 is converted into the form of an output vector 1528 through further processing.

The output vector 1528 is output from the tracklet network through the output layer 1530 .

According to one embodiment, the remote server 130 includes a second machine learning model. The second machine learning model included in the remote server 130 includes a tracklet network 1500 . The first machine learning model includes a structure in which at least one bypass path 1540 is added to the structure of the tracklet network 1500 . The start and end points of the bypass path 1540 and the number of bypass paths 1540 may vary according to embodiments. When the bypass path 1540 is disposed, the output value of the layer corresponding to the start point of the bypass path is transmitted to the layer corresponding to the end point of the bypass path, Processing is not performed and is bypassed. The first machine learning model has the effect of reducing the throughput and remarkably improving the processing speed by applying the bypass path 1540 to the tracklet network 1500 of the second machine learning model.

The remote server 130 receives an input image from the edge computing device 110 and generates vehicle information and speed information using the second machine learning model. According to an embodiment, the edge computing device 110 transmits the input frame section of the input image in which the speeding event is detected to the remote server 130 , and the remote server 130 receives the input received from the edge computing device 110 . The frame section is input to the second machine learning model, vehicle information and speed information are generated and transmitted to the edge computing device 110 . According to another embodiment, the edge computing device 110 transmits the input image and the speeding event information together to the remote server 130 , and the remote server 130 provides vehicle information and speed for an input frame section in which the speeding event is detected. Information is generated and transmitted to the edge computing device 110 . When the edge computing device 110 receives vehicle information and speed information from the remote server 130 , the edge computing device 110 updates vehicle information and speed information generated by the edge computing device 110 with the received vehicle information and speed information.

Referring back to FIG. 9 , the object identification information, speed information, and tracklet information 918 of each input frame generated by the machine learning model processing 920 are post-processed 930 . Post-processing 930 includes graph model generation processing 932 and clustering processing 936 .

16 is a diagram illustrating a graph model according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the graph model generation process 932 generates the graph model 1600 from identification information of an object, velocity information, and tracklet information. The graph model 1600 includes at least one node 1610 and at least one connection line 1620 connecting the nodes 1610 . In each node 1610, information in each frame of the same object is stored. The graph model generation process 932 may identify and store information about the same object in each node based on the identification information of the object generated by the machine learning model. For example, each node 1610 may include at least one or a combination of object region information, speed information, and image data 1612 in each frame for the same object.

A connection line 1620 disposed between each node 1610 indicates a relationship between each object. Relevance is information indicating the similarity of paths between objects. The graph model 1600 may be expressed as g(V,E) which is a function of the tracklet information (V) and the relevance information (E).

17 is a diagram illustrating a process of calculating a relation between objects according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, post-processing 930 includes relevance calculation processing 1700 . The relevance calculation processing 1700 selects the tracklet information of two different objects from the tracklet information and inputs it to the tracklet network 1720 . The tracklet information of the two different selected objects may include tracklet information extracted from a plurality of frames for each object.

The tracklet network 1720 generates relevance information (E) by calculating the similarity of paths for two different objects. When the relevance information E is high, it can be determined that the similarity of the paths of two objects is high, and when the relevance information E is low, it can be determined that the similarity of the paths of the two objects is low. The relevance information E may be defined as a value of 0 or more and 1 or less. If tracklet information for the same object is input to the tracklet network, the relevance information (E) is 1, and if tracklet information of different objects is input to the tracklet network 1720, the relevance information (E) is A value less than 1 can be returned as a value.

18 is a diagram illustrating a graph model generation process and a clustering process according to an embodiment of the present disclosure.

The graph model generation processing 932 performs a graph model generation processing based on the tracklet information and the relevance information E as described above. The graph model generation process 932 may change the property of the connection line between each node based on the relevance information E. For example, the graph model generation process 932 may change the thickness of the connecting line or change the color of the connecting line to display the relevance information E on the connecting line. In addition, the graph model generation process 932 arranges related nodes adjacent to each other and connects each node by the relevance information E, and disposes non-relevant nodes away from each other so that no connecting lines are placed between unrelated nodes. it may not be

The clustering process 934 may cluster the objects based on the relevance information E. For example, the clustering process 934 may define highly related objects as one cluster 1310a, 1310b, and 1310c.

19 is a diagram illustrating a result image according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, identification information and speed information of an object may be inserted into an input frame. The result image 1900 may include a block 1910 indicating an object region, speed information 1920 , and path information 1930 . The object identification information may be indicated by the color of the block 1910 . The result image 1900 may indicate a block 1910 , speed information 1920 , and path information 1930 for each of the at least one detected object. The route information 1930 may indicate route information tracked for a preset time period.

The resulting image 1900 may be generated for all frames of the input video or only for a frame of interest in which a speeding event is detected. The first processor 210 may generate the result image 1900 and output it through an output interface or transmit it to the first remote device 120 .

20 is a diagram showing the structure of a school zone monitoring system according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the school zone monitoring system 100a includes an edge computing device 110 , a school zone monitoring device 2010 , a client device 122 , and a remote server 130 . According to an embodiment, the school zone monitoring device 2010 corresponds to the first remote device 120 described above. According to another embodiment, the school zone monitoring device 2010 may correspond to a device other than the first remote device 120 . For example, the school zone monitoring device 2010 may correspond to an electronic device disposed in a public institution such as a police station or a district office, or may correspond to an electronic device having a predetermined authority. In the present disclosure, an embodiment in which the school zone monitoring device 2010 corresponds to the first remote device 120 described above will be mainly described. However, the exemplary embodiment of the present disclosure is not limited to that the school zone monitoring device 2010 corresponds to the first remote device 120 , and the school zone monitoring device 2010 may be variously changed according to the exemplary embodiment.

The school zone monitoring device 2010 includes a second communication unit 2012 and a second processor 2014 . The second communication unit 2012 corresponds to the second communication unit 222 of the first remote device 120 described with reference to FIG. 2 , and the second processor 2014 is the first remote device 120 described with reference to FIG. 2 . It corresponds to the second processor 220 . Therefore, in the description of FIG. 20, a description overlapping with the previously described content will be omitted.

The second communication unit 2012 receives at least one of an input image, vehicle information, speed information, pedestrian information, and speeding event information or a combination thereof from the edge computing device 110 . The second processor 2014 generates monitoring information by using at least one of an input image received from the edge computing device 110 , vehicle information, speed information, and pedestrian information, or a combination thereof. In addition, the second processor 2014 generates a traffic map guide from the monitoring information.

The second processor 2014 receives at least one of input image, vehicle information, speed information, pedestrian information, monitoring information, traffic map guide, and speeding event information to the remote server 130 through the second communication unit 2012 or these Send the combination. In addition, the second processor 2014 transmits at least one of the monitoring information and the traffic map guide or a combination thereof to the client device 122 through the second communication unit 2012 .

According to one embodiment, the remote server 130 includes a second machine learning model. The first machine learning model of the edge computing device 110 has a structure in which the bypass path 1540 is applied to the tracklet network 1500 of the second machine learning model, as described above with reference to FIG. 15 . When the received speeding event information includes event detection information indicating that a speeding event is detected, the school zone monitoring device 2010 transmits an input image of a section corresponding to the speeding event and speeding event information to the remote server 130 . do.

The remote server 130 receives an input image from the school zone monitoring device 2010 and generates vehicle information and speed information using the second machine learning model. According to an embodiment, the school zone monitoring device 2010 transmits the input frame section of the input image in which the speeding event is detected to the remote server 130 , and the remote server 130 receives it from the school zone monitoring device 2010 . The input frame section is input to the second machine learning model, vehicle information and speed information are generated and transmitted to the school zone monitoring device 2010 . According to another embodiment, the school zone monitoring device 2010 transmits the input image and speeding event information together to the remote server 130, and the remote server 130 provides vehicle information and Speed information is generated and transmitted to the school zone monitoring device 2010 . When the school zone monitoring device 2010 receives vehicle information and speed information from the remote server 130, it is generated in the edge computing device 110 with the received vehicle information and speed information and updates the received vehicle information and speed information. .

21 is a flowchart illustrating a school zone monitoring method according to an embodiment of the present disclosure.

Each step of the school zone monitoring method according to an embodiment of the present disclosure may be performed by various types of electronic devices including a processor and a communication unit. The present disclosure will focus on an embodiment in which the school zone monitoring apparatus 2010 according to embodiments of the present disclosure performs a school zone monitoring method. Therefore, the embodiments described for the school zone monitoring device 2010 are applicable to the embodiments for the school zone monitoring method, on the contrary, the embodiments described for the edge computing device control method is the school zone monitoring device (2010) Applicable to the embodiments. The school zone monitoring method according to the disclosed embodiments is not limited to being performed by the school zone monitoring device 2010 disclosed in the present disclosure, and may be performed by various types of electronic devices.

First, the edge computing device 110 collects an input image, and transmits the collected input image to the school zone monitoring device 2010 (S2102). The school zone monitoring device 2010 receives an input image from the edge computing device 110 . In addition, the edge computing device 110 uses the first machine learning model from the input image to generate at least one or a combination of vehicle information, speed information, pedestrian information, and speeding event information, and a school zone monitoring device (2010) to (S2104). The school zone monitoring device 2010 receives at least one or a combination of vehicle information, speed information, pedestrian information, and speeding event information from the edge computing device 110 .

The school zone monitoring device 2010 generates monitoring information based on at least one or a combination of the received vehicle information, speed information, pedestrian information, and speeding event information (S2106). Also, the school zone monitoring device 2010 may generate a traffic map guide based on the monitoring information.

The school zone monitoring device 2010 transmits the generated monitoring information to the client device (S2108). In addition, the school zone monitoring device 2010 transmits a traffic map guide to the client device 122 .

In addition, the school zone monitoring device 2010 transmits at least one or a combination of the input image, vehicle information, speed information, and pedestrian information collected to the remote server 130 to the remote server 130 ( S2110 ). The school zone monitoring device 2010 may also transmit information such as generated monitoring information and a traffic map guide to the remote server 130 .

According to an embodiment, when a speeding event is detected, the school zone monitoring apparatus 2010 transmits an input image of a section in which the speeding event is detected and speeding event information to the remote server 130 . The remote server 130 uses the second machine learning model in which some bypass paths are removed from the first machine learning model of the edge computing device 110 for the input image of the section in which the speeding event is detected, vehicle information, The speed, information, and pedestrian information are generated and transmitted to the school zone monitoring device 2010 (S2112).

22 is a diagram illustrating a process of generating school zone risk information and a traffic map guide according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the school zone monitoring apparatus 2010 generates school zone risk information and a traffic map guide from the input image 2220 . First, the school zone monitoring device 2010 calculates the number of vehicles and pedestrians included in the input image 2220 from the vehicle information and pedestrian information received from the edge computing device 110 (S2202). The vehicle information includes information on an area corresponding to a vehicle in the input image 2220, and the school zone monitoring apparatus 2010 may calculate the number of vehicles by counting the number of areas corresponding to the vehicle from the vehicle information. The pedestrian information includes information on areas corresponding to pedestrians in the input image 2220, and the school zone monitoring apparatus 2010 may calculate the number of vehicles by counting the number of areas corresponding to pedestrians from the pedestrian information.

Next, the school zone monitoring apparatus 2010 determines whether the number of vehicles exceeds a vehicle number reference value (S2204). In addition, the school zone monitoring apparatus 2010 determines whether the number of pedestrians exceeds a reference value for the number of pedestrians (S2206). The school zone monitoring apparatus 2010 calculates school zone risk information based on the result of determining whether the reference values for the number of vehicles and the number of pedestrians are exceeded (S2208). The school zone monitoring device 2010 increases the risk level when the number of vehicles exceeds the reference value for the number of vehicles, and increases the level of risk when the number of pedestrians exceeds the reference value for the number of pedestrians.

In addition, the school zone monitoring device 2010 generates a traffic map guide based on the calculated school zone risk information (S2210). The traffic guidance guide includes, for example, a first-level guide that requests guidance from a tutor to provide guidance to and from school, and a second-level guide that requests a traffic guidance from a traffic police. The school zone monitoring device 2010 generates, for example, a second level guide when the number of vehicles exceeds a reference value, and a first level guide when the number of vehicles does not exceed the reference value and the number of pedestrians exceeds the reference value. to create

23 is a view for explaining a process of generating school zone risk information and a traffic map guide from school zone environment information, according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the school zone monitoring device 2010 collects various types of school zone environment information 2310, and based on the collected school zone environment information 2310, school zone risk information 2320 ) is created. The school zone environment information 2310 includes at least one or a combination of the number of vehicles, the number of pedestrians, weather information, time information, and the number of parked and stopped vehicles.

As described above with reference to FIG. 22 , the vehicle number information and pedestrian number information may be calculated from vehicle information and pedestrian information received from the edge computing device 110 .

The weather information may be obtained from an external device or an external server. For example, the school zone monitoring device 2010 receives weather information from the Meteorological Administration server or a weather information providing service server. The weather information may have a predetermined value, such as rain, sunny, snow, and fog.

The time information may be obtained from time information held by the school zone monitoring device 2010 . The time information may be defined in a predetermined time zone. For example, the time information may have a predetermined value, such as a start time, a dismissal time, a work time, a leave time, a congestion time, and a normal time.

The number of parked and parked vehicles means the number of parked and parked vehicles detected in the school zone. Based on the speed information received from the edge computing device 110 , the school zone monitoring device 2010 defines a vehicle in which the speed information is 0 or a value close to 0 and is maintained for more than the parking and stopping reference time as a parked and stopped vehicle. In addition, the school zone monitoring apparatus 2010 defines a vehicle detected in the outermost lane or a vehicle adjacent to a sidewalk as the vehicle in which the speed information is maintained at 0 or an approximate value of 0 for more than the standard time for parking and stopping, as a vehicle for parking and stopping. The parking stop reference time may be set longer than, for example, a signal waiting time indicating an interval between driving signals of vehicle signals.

The school zone monitoring device 2010 generates school zone risk information 2320 based on the collected or calculated school zone environment information 2310 . The risk level of the school zone risk information 2320 is determined based on various school zone environment information 2310 . The risk level is defined as a predetermined number.

The school zone monitoring apparatus 2010 highly evaluates the risk level when a predetermined reference value is exceeded for each of the number of vehicles, the number of pedestrians, or the number of parked and stopped vehicles. In addition, the school zone monitoring device 2310 evaluates the risk level high when the weather information corresponds to rain, snow, or fog. In addition, the school zone monitoring device 2310 evaluates the risk level to be high in case of going to school time, leaving school time, going to work time, leaving time, or congested time.

According to an embodiment, the school zone monitoring apparatus 2010 evaluates the risk level to be high when any one of the school zone environment information 2310 corresponds to a condition of a high risk level. For example, when the weather information is rainy, the school zone risk information 2320 may be evaluated as a high risk level even if the other school zone environment information 2310 is not evaluated as a high risk level.

According to another embodiment, the school zone monitoring device 2010 is based on the school zone environment information 2310 only when the time information is the start time or the dismissal time, the school zone risk information 2320 and the traffic map guide 2330. , and if the time information is not the start time or the dismissal time, the school zone risk information 2320 and the traffic map guide 2330 are not evaluated.

According to another embodiment, the school zone monitoring device 2010 calculates a weight for each school zone environment information 2310, and calculates a weighted average evaluation value for the progress of judgment for each school zone environment information 2310 and calculates a risk level based on the evaluation value.

The school zone monitoring device 2010 generates a traffic map guide 2330 based on the school zone risk information 2320 . The traffic map guide 2330 may be defined at various levels according to embodiments. For example, the traffic map guide 2330 may include a first-level guide that requests a teacher's guidance to and from school and a second-level guide that requests a traffic police's guidance to and from school.

Meanwhile, the disclosed embodiments may be implemented in the form of a computer-readable recording medium storing instructions and data executable by a computer. The instructions may be stored in the form of program code, and when executed by the processor, a predetermined program module may be generated to perform a predetermined operation. Further, the instruction, when executed by a processor, may perform certain operations of the disclosed embodiments.

As described above, the disclosed embodiments have been described with reference to the accompanying drawings. Those of ordinary skill in the art to which the present invention pertains will understand that the present invention may be practiced in other forms than the disclosed embodiments without changing the technical spirit or essential features of the present invention. The disclosed embodiments are illustrative and should not be construed as limiting.

100, 100a School Zone Monitoring System
110, 110a Edge Computing Device
120 first remote device
122 client device
130 remote server
140 second remote device
150 third remote device
210, 210a, 210b first processor
212 first machine learning model
220 second processor

Claims (18)

an input interface for receiving an input image photographed from at least one camera for photographing a school zone;
a video buffer for storing the input image;
The input image captured by the at least one camera is input to a first machine learning model, and vehicle information and speed information detected from the input image are obtained from the first machine learning model, and the detected speed information is a speed reference. at least one first processor to detect a speeding event exceeding a value; and
A first remote device installed in a school, and a first communication unit communicating with a second remote device installed to be spaced apart from the installation point of the camera in a direction opposite to the vehicle traveling direction by a first reference distance,
The at least one first processor, to the first remote device and the second remote device, includes at least one of the vehicle information, the speed information, speeding event information for the speeding event, or the input image, or a combination thereof Control the first communication unit to transmit
The first machine learning model generates vehicle area information for each of one or a plurality of vehicles recognized in the input image,
the at least one first processor transmits the input image and the vehicle area information for each of one or a plurality of vehicles recognized in the input image to the first remote device;
The first remote device calculates the number of vehicles included in the input image by counting the number of vehicle areas detected in the input image, and generates school zone risk information based on the number of vehicles. According to claim 1,
The at least one first processor transmits the input image, the vehicle information, and the speed information to the first remote device, and transmits the vehicle information, the speed information, and the speed information to the second remote device. An edge computing device that controls the first communication unit to transmit event information. According to claim 1,
The at least one first processor obtains, from the first machine learning model, information on the number of pedestrians in the school zone,
An edge computing device for controlling the first communication unit to transmit the number of pedestrians information to at least one of the first remote device and the second remote device. 4. The method of claim 3,
The first communication unit communicates with a third remote device that provides pedestrian information,
The at least one first processor is configured to transmit, to the third remote device, the pedestrian information. According to claim 1,
The first communication unit includes a BCDMA communication module, and transmits the vehicle information, the speed information, and the speeding event information through the serial port of the BCDMA communication module, and transmits the input image through an Ethernet port. edge computing device. According to claim 1,
The at least one first processor detects the license plate area and the driver area of the vehicle detected from the input image, and performs image processing to reduce visibility of the license plate area and the driver area to remove personal information from the corrected input image and to transmit the corrected input image through the first communication unit. The method of claim 1,
The at least one first processor, based on the vehicle information and the speed information obtained from the first machine learning model, calculates the traffic information of the school zone, and transmits the traffic information through the first communication unit. An edge computing device that transmits to the first remote device or the second remote device. According to claim 1,
The at least one camera is installed in each lane of the school zone,
The at least one first processor is configured to obtain vehicle information and speed information for each lane based on an input image for each lane obtained from the at least one camera. According to claim 1,
When a speeding event is detected through the first communication unit, the at least one first processor transmits the speeding event information and the input image corresponding to the detected speeding event to a remote server, and from the remote server Receive speed information, vehicle information, and speeding event information detected for the input image,
The remote server obtains the speed information and the vehicle information from the input image using a second machine learning model,
The first machine learning model is a machine learning model to which a bypass path is applied to at least one layer of the second machine learning model, an edge computing device. Receiving an input image photographed from at least one camera for photographing the school zone;
storing the input image;
inputting the input image captured by the at least one camera into a first machine learning model, and obtaining vehicle information and speed information detected from the input image from the first machine learning model;
detecting a speeding event in which the detected speed information exceeds a speed reference value; and
A first remote device installed in a school and a second remote device installed to be spaced apart from the installation point of the camera by a first reference distance in a direction opposite to the vehicle traveling direction, the vehicle information, the speed information, and a speeding event for the speeding event transmitting information or at least one of the input images or a combination thereof; and
including,
The first machine learning model generates vehicle area information for each of one or a plurality of vehicles recognized in the input image,
The transmitting includes transmitting the input image and the vehicle area information for each of one or a plurality of vehicles recognized in the input image to the first remote device,
The first remote device calculates the number of vehicles included in the input image by counting the number of vehicle areas detected in the input image, and generates school zone risk information based on the number of vehicles, Edge computing device control Way. In the school zone monitoring method performed by the school zone monitoring device,
Receiving an input image captured by at least one camera installed in the school zone;
receiving vehicle information, speed information, and pedestrian information detected from the input image from a remote edge computing device;
generating monitoring information for the school zone based on the input image, the vehicle information, the speed information, and the pedestrian information;
transmitting the monitoring information to at least one client device; and
Transmitting at least one of the input image, the vehicle information, the speed information, the pedestrian information, and the monitoring information to a remote server,
The monitoring information includes school zone risk information,
calculating the number of vehicles in the school zone;
calculating the school zone risk information based on whether the number of vehicles in the school zone exceeds a vehicle number reference value; and
Further comprising the step of generating and outputting a traffic map guide based on the school zone risk information, school zone monitoring method. 12. The method of claim 11,
The pedestrian information includes information on the number of pedestrians,
The calculating of the school zone risk information may include: based on whether the number of vehicles in the school zone exceeds the reference value for the number of vehicles and whether the number of pedestrians in the school zone exceeds the reference value for the number of pedestrians, the school zone risk information Including the step of calculating, school zone monitoring method. 13. The method of claim 12,
The traffic map guide includes a first level guide for requesting a teacher's guidance to and from school, and a second level guide for requesting a traffic map from the traffic police, a school zone monitoring method. 13. The method of claim 12,
The method is
collecting weather information; and
further comprising collecting time information;
In the calculating of the school zone risk information, the school zone monitoring method is further based on the weather information and the time information, calculating the school zone risk information. 13. The method of claim 12,
The method is
detecting the number of parked and stopped vehicles in the outermost lane; and
The method of monitoring a school zone, further comprising the step of calculating the school zone risk information based on whether the number of parked and stopped vehicles exceeds a parked and stopped reference value. 12. The method of claim 11,
The edge computing device includes a first machine learning model that generates the vehicle information and the speed information from the input image,
The method is
receiving, from the edge computing device, speeding event information for detecting a vehicle having a speed exceeding a speed reference value;
transmitting at least one input image in which the speeding event information is detected to the remote server;
receiving speed information, vehicle information, and speeding event information detected with respect to the input image from the remote server; and
The method further comprising the step of updating the monitoring information with the speed information, the vehicle information, and the speeding event information received from the remote server,
The remote server obtains the speed information and the vehicle information from the input image using a second machine learning model,
The first machine learning model is a machine learning model to which a bypass path is applied to at least one layer of the second machine learning model, a school zone monitoring method. a communication unit that receives an input image from at least one camera installed in a school zone, and receives vehicle information, speed information, and pedestrian information detected from the input image from a remote edge computing device; and
Based on the input image, the vehicle information, the speed information, and the pedestrian information, the monitoring information for the school zone is generated, and the monitoring information is transmitted to at least one client terminal through the communication unit, and the communication unit A processor for transmitting at least one of the input image, the vehicle information, the speed information, the pedestrian information, and the monitoring information to a remote server through
The monitoring information includes school zone risk information,
The processor is
Calculating the number of vehicles in the school zone,
Based on whether the number of vehicles in the school zone exceeds a vehicle number reference value, calculating the school zone risk information,
A school zone monitoring device for generating and outputting a traffic map guide based on the school zone risk information. 18. The method of claim 17,
The edge computing device includes a first machine learning model that generates the vehicle information and the speed information from the input image,
The processor receives, from the edge computing device, speeding event information in which a vehicle having a speed exceeding a speed reference value is detected, and transmits at least one input image in which the speeding event information is detected to the remote server, , receives the speed information, pedestrian information, and speeding event information detected for the input image from the remote server, and receives the monitoring information with the speed information, the vehicle information, and the speeding event information received from the remote server update,
The remote server obtains the speed information and the vehicle information from the input image using a second machine learning model,
The first machine learning model is a machine learning model to which a bypass path is applied to at least one layer of the second machine learning model, a school zone monitoring device.

Images