WO2019223655A1

WO2019223655A1 - Detection of non-motor vehicle carrying passenger

Info

Publication number: WO2019223655A1
Application number: PCT/CN2019/087648
Authority: WO
Inventors: 龙传书
Original assignee: 杭州海康威视数字技术股份有限公司
Priority date: 2018-05-22
Filing date: 2019-05-20
Publication date: 2019-11-28
Also published as: CN110516518A

Abstract

Provided are a method and apparatus for detecting a non-motor vehicle carrying a passenger, and an electronic device. As an example, the method for detecting a non-motor vehicle carrying a passenger comprises: acquiring a target image to be detected; performing non-motor vehicle detection on the target image to obtain a non-motor vehicle area in the target image, wherein the non-motor vehicle area contains information of a non-motor vehicle; and detecting whether the non-motor vehicle in the non-motor vehicle area carries a passenger, to obtain a detection result of the non-motor vehicle area.

Description

Non-motorized vehicle detection

Technical field

The present application relates to the field of intelligent transportation, and in particular, to a method, device, and electronic device for detecting non-motorized vehicle-borne persons.

Background technique

In order to avoid the difficulties of crowding buses, subways, and traffic congestion, when traveling short distances, people are more inclined to choose non-motorized vehicles, such as bicycles, battery cars, mopeds, tricycles, etc.

In road traffic, due to the lack of protective equipment, people using non-motorized vehicles are the weakest among road traffic participants, and they are most vulnerable when they participate in traffic activities. And the use of non-motorized vehicles is more likely to cause people to be harmed. According to statistics, many serious non-motor vehicle accidents currently occur are related to people.

In order to help monitor the behavior of non-motorized vehicles and protect people's travel safety, how to quickly and effectively detect whether non-motorized vehicles carry people is an urgent problem.

Summary of the Invention

In view of this, embodiments of the present application provide a non-motorized vehicle-borne person detection method, device, and electronic device to quickly and effectively detect whether a non-motorized vehicle carries a person.

In a first aspect, an embodiment of the present application provides a method for detecting a non-motorized vehicle person, including: obtaining a target image to be detected; performing non-motor vehicle detection on the target image to obtain a non-motor vehicle in the target image; A motor vehicle area; wherein the non-motor vehicle area contains information of a non-motor vehicle; detecting whether a non-motor vehicle in the non-motor vehicle area carries a person, and obtaining a detection result of the non-motor vehicle area.

In a second aspect, an embodiment of the present application further provides a non-motor vehicle-mounted person detection device, including: an image obtaining unit, a non-motor vehicle area obtaining unit, and a person carrying detection unit. The image obtaining unit is configured to obtain a target image to be detected. The non-motor vehicle region obtaining unit is configured to perform non-motor vehicle detection on the target image to obtain a non-motor vehicle region in the target image; wherein the non-motor vehicle region includes information of a non-motor vehicle. The human-carrying detection unit is configured to detect whether a non-motorized vehicle in the non-motorized area carries a person, and obtain a detection result of the non-motorized area.

In a third aspect, an embodiment of the present application further provides an electronic device including an internal bus, a memory, a processor, and a communication interface. The processor, the communication interface, and the memory communicate with each other through the internal bus; the memory is used to store machine feasible instructions corresponding to the non-motorized vehicle detection method; and the processor is used to Read the machine-readable instructions on the memory and execute the instructions to implement the non-motor vehicle-mounted person detection method provided in the embodiment of the present application.

According to a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the computer program is processed by a processor, the non-motorized vehicle-borne person detection method provided by the embodiment of the present application is implemented. .

In the method provided in the embodiment of the present application, non-motor vehicle detection is performed on the obtained target image to obtain a non-motor vehicle area in the target image; detecting whether a non-motor vehicle in the non-motor area carries a person, and obtaining the non-motor vehicle. Detection results corresponding to the motor vehicle area. It can be seen that this solution first identifies the non-motorized vehicle area and then identifies whether the non-motorized vehicle carries people in the non-motorized vehicle area, so that the identification of the human carrying is well targeted, and therefore, it can quickly and effectively detect whether the non-motorized vehicle carries people.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a flowchart of a non-motorized vehicle detection method provided by an embodiment of the present application;

Figures 2 (a), (b), and (c) are schematic diagrams of interfaces marked with non-motor vehicle areas and their detection results;

FIG. 3 is another flowchart of a non-motorized vehicle detection method provided by an embodiment of the present application; FIG.

4 is a schematic structural diagram of a non-motorized vehicle-mounted person detection device according to an embodiment of the present application;

FIG. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed ways

Exemplary embodiments will be described in detail here, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numerals in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with this application. Rather, they are merely examples of devices and methods consistent with certain aspects of the application as detailed in the appended claims.

The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this application and the appended claims, the singular forms "a", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

It should be understood that although the terms first, second, third, etc. may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein can be interpreted as "at" or "when" or "in response to determination".

In order to quickly and effectively detect whether a non-motor vehicle carries a person, embodiments of the present application provide a non-motor vehicle-borne person detection method, device and electronic device.

The following first introduces a non-motorized vehicle-borne person detection method provided by an embodiment of the present application.

It should be noted that the execution subject of the non-motorized vehicle-borne person detection method provided in the embodiments of the present application may be a non-motorized vehicle-borne person detection device. In specific applications, the non-motorized vehicle-mounted person detection device can run on a terminal device or a server, which is reasonable. For the convenience of reference, this article simplifies the "non-motor vehicle-mounted person detection device" from its name to "detection device" for reference.

In addition, the so-called non-motor vehicle may be a bicycle, a battery car, a moped, a tricycle, or the like.

As shown in FIG. 1, a non-motorized vehicle-borne person detection method provided in an embodiment of the present application may include the following steps S101 to S103.

S101. Obtain a target image to be detected.

Among them, since non-motorized vehicle detection is used as a means to assist road monitoring, the target image can be an image captured by a camera installed on the road, or a video in a video collected by a surveillance camera installed on the road Frames, of course, are not limited to this.

S102. Perform non-motor vehicle detection on the target image to obtain a non-motor vehicle area in the target image.

After obtaining the target image, the detection device may directly perform non-motor vehicle detection on the target image to obtain a non-motor vehicle region in the target image; wherein the non-motor vehicle region contains information of the non-motor vehicle. In addition, the detection device can detect one non-motor vehicle area or at least two non-motor vehicle areas in one target image. For each non-motor vehicle area, the steps of S103 can be performed to obtain each Test results of non-motor vehicles carrying people in the non-motor vehicle area.

In one embodiment, because the target image is an image about a road scene, the target image usually has noise interference, and the images collected by different acquisition devices may have very different imaging characteristics, such as resolution, size, etc. These all have a certain impact on the detection process. Therefore, in order to eliminate these effects, after obtaining the target image, the detection device may perform image preprocessing on the target image, and then perform non-motor vehicle detection on the target image that has undergone image preprocessing. The image preprocessing may include at least one of denoising, histogram equalization, and size normalization, which is not limited to this.

In addition, in the monitoring scene of a capture device such as a snap camera or a surveillance camera, the area where the vehicle is running can be considered to be in a fixed position. At this time, the images collected by the capture device are required for the non-motorized vehicle detection process. The effective detection area is fixed. For example, the surveillance scene of a surveillance camera includes urban roads and green belts on both sides of urban roads. The scene area where vehicles are running is urban roads. Therefore, in the images collected by this surveillance camera, The part about the urban road is a valid detection area, and the part about the green belt is an invalid detection area, and the position of the area of the urban road part in the image is fixed. For the case where the effective detection area in the image is at a fixed position, in order to reduce the amount of data calculation and improve the detection accuracy, step S102 is a step of performing non-motor vehicle detection on the target image to obtain the non-motor vehicle area in the target image. It may include: determining a valid detection area from the target image based on preset area calibration information; performing non-motor vehicle detection on the valid detection area to obtain a non-motor vehicle area in the target image.

The method for setting the area calibration information may include one of the following methods: Method 1. For scenes in which images captured by a snapshot camera or a monitoring camera are displayed in real time through a monitoring screen, the area calibration may be set by delimiting the area on the monitoring screen. Information; mode two, setting the area calibration information by giving coordinate information; mode three, the system automatically sets the area calibration information according to the default value. And, in specific applications, the image pre-processing and the effective detection area detection process can be used in combination to improve the recognition effectiveness.

In one embodiment, non-motor vehicle detection may be performed on a target image or an effective detection area based on a pre-trained non-motor vehicle detection model, and the non-motor vehicle area in the target image is obtained, of course, it is not limited to this. Among them, the model types of the non-motor vehicle detection model may include, but are not limited to: CNN (Convolutional Neural Network, Convolutional Neural Network), RNN (Recurrent Neural Networks, Recurrent Neural Network), DNN (Deep Neural Network, Deep Neural Network) ,and many more.

In addition, in one embodiment, when training a non-motor vehicle detection model, the non-motor vehicle detection model may be trained to: not only identify a non-motor vehicle area, but also identify a non-motor vehicle type in the non-motor vehicle area category. For the case where the vehicle type can be identified, in order to improve the detection efficiency, after S102 and before S103, the method provided in the embodiment of the present application may further include: obtaining a vehicle type category of the non-motor vehicle area in the non-motor vehicle area; When the preset category is not met, the step of detecting whether a non-motor vehicle in the non-motor vehicle area carries a person in S103 is performed. The preset category includes at least one manned category. For example, the preset categories include tricycles and tandem bicycles.

In one example, the manned situation can be specifically divided into no manned, normal manned (that is, manned) and abnormal manned (that is, not allowed). In the case that the type of the vehicle model meets the preset category, it is determined whether the non-motor vehicle is carrying a person in the non-motor vehicle area, and the detection result is one of no-carrying and normal-carrying. When the type of the vehicle model does not meet the preset category, it is detected whether the non-motor vehicle in the non-motor vehicle area carries a person, and the detection result is one of non-carrying and abnormally carrying a person.

S103: Detect whether a non-motor vehicle in the non-motor vehicle area carries a person, and obtain a detection result of the non-motor vehicle area.

After obtaining the non-motor vehicle area, the detection device can directly detect whether the non-motor vehicle in the non-motor vehicle area is carrying a person, and obtain a detection result corresponding to the non-motor vehicle area. Considering that the non-motor vehicle area contained in the non-motor vehicle area detected by S102 may be incomplete, or fail to completely include passengers and / or drivers, the detection device may detect the non-motor vehicle area. Expand processing to improve the integrity of the recognition area corresponding to the manned detection. Based on the above requirements, in a specific implementation manner, the step of detecting whether a non-motor vehicle in the non-motor vehicle area carries a person and obtaining a detection result of the non-motor vehicle area may include: the non-motor vehicle area The area is subjected to area expansion processing to obtain an expanded area; detecting whether a non-motor vehicle in the expanded area carries a person, and obtaining a detection result corresponding to the non-motor vehicle area.

The non-motorized vehicle area is expanded according to a certain expansion rate. The expansion rate can be set according to experience values. For example, the expansion rate can be 10%, 15%, 20%, 25%, 30%, 40 %and many more.

It should be emphasized that, based on a pre-trained human detection model, it is possible to detect whether a non-motor vehicle in a non-motor vehicle area or an expanded area carries a person, and obtain a detection result for the non-motor vehicle area, of course, it is not limited to this For example, a specific body recognition algorithm can be used to identify the number of people in the non-motor vehicle area or the expanded area, and based on the obtained number, determine the detection result corresponding to the non-motor vehicle area. Among them, the model type of the human detection model may include, but is not limited to: CNN (Convolutional Neural Network, Convolutional Neural Network), RNN (Recurrent Neural Networks, Recurrent Neural Network), DNN (Deep Neural Network, Deep Neural Network), SVM (Support Vector Machine). Among them, SVM is a discriminative method. In the field of machine learning, it is a supervised learning model, which is usually used for pattern recognition, classification, and regression analysis.

In addition, after the detection result of the non-motor vehicle area is obtained, the non-motor vehicle area and the detection result thereof can be marked in the target image, so as to be reflected in an intuitive form. The specific labeling form may include, but is not limited to, a combination of a labeling frame and a text. For a combination of a labeling frame and a text, see FIG. 2 (a), FIG. 2 (b), and FIG. 2 (c). Given interface example.

In the method provided in the embodiment of the present application, non-motor vehicle detection is performed on the obtained target image to obtain a non-motor vehicle area in the target image; and whether the non-motor vehicle in the non-motor vehicle area is carrying a person is obtained. The detection result corresponding to the non-motor vehicle area. It can be seen that this solution first identifies the non-motorized vehicle area and then identifies whether the non-motorized vehicle carries people in the non-motorized vehicle area, so that the identification of the human carrying is well targeted, and therefore, it can quickly and effectively detect whether the non-motorized vehicle carries people.

The following describes in detail a non-motor vehicle-borne person detection method provided in the embodiments of the present application in combination with specific embodiments.

In the embodiment of the present application, the so-called non-motor vehicle may be a bicycle, a battery car, a moped, a tricycle, or the like.

As shown in FIG. 3, a non-motorized vehicle-borne person detection method provided in an embodiment of the present application may include the following steps S201 to S203.

S201. Obtain a target image to be detected.

S202. Based on a pre-trained non-motor vehicle detection model, non-motor vehicle detection is performed on the target image to obtain a non-motor vehicle area.

After obtaining the target image, the detection device may directly perform non-motor vehicle detection on the target image based on a pre-trained non-motor vehicle detection model to obtain a non-motor vehicle area; wherein the non-motor vehicle area includes a non-motor vehicle information. In addition, the detection device can detect one non-motor vehicle area or at least two non-motor vehicle areas in one target image. The steps of S203 can be performed for each non-motor vehicle area to obtain each non-motor vehicle area. Detection results corresponding to the motor vehicle area.

It can be understood that after obtaining the target image, the detection device can perform image preprocessing on the target image, and then perform non-motor vehicle detection on the target image after image preprocessing to eliminate noise interference and different acquisitions. The effect of different imaging characteristics of the images acquired by the device on the detection process. The image preprocessing may include at least one of denoising, histogram equalization, and size normalization, which is not limited to this. In addition, for a case where the effective area in the image is at a fixed position, the step of performing non-motor vehicle detection on the target image based on a pre-trained non-motor vehicle detection model to obtain the non-motor vehicle area in the target image may include: : Determine the effective detection area from the target image based on the preset area calibration information; perform non-motor vehicle detection on the effective detection area based on the pre-trained non-motor vehicle detection model to obtain the non-motor vehicle area in the target image .

It should be emphasized that the model types of the non-motor vehicle detection models involved in this application may include, but are not limited to: CNN (Convolutional Neural Network, Convolutional Neural Network), RNN (Recurrent Neural Networks, Recurrent Neural Network), DNN ( Deep Neural Network), and so on. The specific model training process may be the same as the model training process in related technologies, and details are not described herein.

It can be understood that the image samples used to train the non-motor vehicle detection model can be extracted from the surveillance videos of the downtown area and highway intersections, etc., of course, it is not limited to this. In addition, the image samples used to train the non-motor vehicle detection model may include only positive samples, and may include both positive samples and negative samples. Among them, the positive samples may include images of non-motor vehicles, and the negative samples may include vehicle information. And / or images of motor vehicles and non-motor vehicles. In addition, before training the non-motor vehicle detection model, the non-motor vehicle area in the image sample can be calibrated, that is, the non-motor vehicle area is marked by a rectangular frame, and the model training program can obtain the calibrated non-motor vehicle area. And extract the non-motor vehicle area based on the coordinate information and then perform model training.

Since the purpose of the embodiment of the present application is to detect whether a non-motor vehicle carries a person, when the non-motor vehicle area is calibrated, both the driver and the passenger can be framed in the non-motor vehicle area.

In one embodiment, when training a non-motor vehicle detection model, the non-motor vehicle detection model may be trained to: not only identify a non-motor vehicle area, but also identify a type of a non-motor vehicle in the non-motor vehicle area. For the case where the vehicle type can be identified, in order to improve the detection efficiency, after S202 and before S203, the method provided in the embodiment of the present application may further include: obtaining a vehicle type category of the non-motor vehicle area in the non-motor vehicle area; When the preset category is not met, S203 is executed.

S203. Based on a pre-trained human-carrying detection model, it is detected whether a non-motorized vehicle in the non-motorized vehicle area carries a person, and a detection result of the non-motorized vehicle area is obtained.

After the non-motor vehicle area is obtained, the detection device may directly detect whether a non-motor vehicle in the non-motor vehicle area is carrying a person based on a pre-trained human-carrying detection model, and obtain a detection result corresponding to the non-motor vehicle area. Considering that the non-motor vehicle area contained in the non-motor vehicle area detected by S202 may be incomplete or fail to include passengers completely, the detection device can expand the non-motor vehicle area, thereby Improve the integrity of the recognition area corresponding to human detection. Based on the above requirements, in a specific implementation manner, the detection of a non-motor vehicle in the non-motor vehicle area based on a pre-trained human-carrying detection model is performed to obtain a detection result corresponding to the non-motor vehicle area. The steps may include: performing an area expansion process on the non-motorized vehicle area to obtain an expanded area; and detecting whether a non-motorized vehicle in the expanded area is carrying a person based on a pre-trained human detection model, and obtaining the non-motorized vehicle area location. Corresponding test results.

Optionally, the step of detecting whether a non-motor vehicle in the non-motor vehicle area is carrying a person based on a pre-trained human-carrying detection model, and obtaining the detection result of the non-motor vehicle area may include: The manned detection model determines the manned confidence level of the non-motorized vehicle in the non-motorized vehicle area; when the manned confidence level is greater than a preset confidence threshold, it is determined that the detection result of the non-motorized vehicle area is manned.

Among them, the range of manned confidence is [0,1], which is not limited to this, of course.

In one example, the manned situation can be specifically divided into no manned, normal manned and abnormal manned. When the manned confidence level is greater than a pre-set reliability threshold value, it is determined that the detection result of the non-motor vehicle area is abnormal or normal. When the manned confidence level is not greater than a preset confidence threshold, it is determined that the detection result corresponding to the non-motorized vehicle area is no manned. The preset reliability threshold can be set according to the actual situation. For example, when the value range of the human confidence is [0,1], the preset reliability threshold can be 0.6, 0.7, 0.8, 0.9, etc. Wait.

Optionally, the step of detecting whether a non-motor vehicle in the non-motor vehicle area is carrying a person based on a pre-trained human-carrying detection model, and obtaining the detection result of the non-motor vehicle area may include: based on the pre-training To determine the manned confidence level of the non-motorized vehicle in the non-motorized vehicle area; determine whether the manned confidence level is greater than a preset confidence threshold; when the determination result is yes, determine the non-motorized vehicle area The corresponding detection result is manned; when the determination result is no, determine whether the number of persons in the non-motorized vehicle area is greater than 1, and if so, determine that the detection result corresponding to the non-motorized vehicle area is manned, otherwise, It is determined that the detection result corresponding to the non-motorized vehicle area is that there is no passenger.

In one example, in the case of manned situations, which are specifically divided into no manned, normal manned, and abnormal manned, when the manned confidence is greater than a pre-set confidence threshold, it is determined that the detection result of the non-motorized vehicle area is Abnormally manned or normally manned. When the manned confidence level is not greater than the preset confidence threshold, determine whether the number of people in the non-motorized vehicle area is greater than 1. If the number of people is greater than 1, determine that the detection result corresponding to the non-motorized vehicle area is abnormally manned or normal. Manned; otherwise, it is determined that the detection result corresponding to the non-motor vehicle area is no manned.

Among them, the number of people in the non-motor vehicle area can be identified by a specific human recognition algorithm. In addition, when it is determined that the number of persons in the non-motor vehicle area is greater than a threshold value such as 1, it indicates that other persons are included in addition to the driver. In another embodiment, based on the pre-trained human-carrying detection model, detecting whether a non-motor vehicle in the non-motor vehicle area carries a person, and obtaining a detection result of the non-motor vehicle area, further including: judging Whether the number of human areas in the non-motor vehicle area is greater than 1; if the number of human areas in the non-motor vehicle area is greater than 1, determining that the detection result of the non-motor vehicle area is a manned person.

In addition, it should be emphasized that the types of human-carrying detection models involved in this application may include, but are not limited to: CNN (Convolutional Neural Network, Convolutional Neural Network), RNN (Recurrent Neural Networks, Recurrent Neural Network), DNN ( Deep Neural Network), SVM, etc. The specific model training process may be the same as the model training process in related technologies, and details are not described herein.

In addition, you can use the screenshot tool to intercept the non-motor vehicle area calibrated by the positive sample of the non-motor vehicle detection model, and then classify the manned situation in the non-motor vehicle area, that is, to give the manned confidence, based on the classification The completed non-motorized vehicle area is trained with a human detection model. It should be noted that, in order to improve fault tolerance, the negative samples of the non-motor vehicle detection model and the false detection samples corresponding to the non-motor vehicle detection model can be used as samples of the human detection model. The manned confidence level of the negative sample of the vehicle detection model, and the manned confidence level of the false detection sample corresponding to the non-motorized vehicle detection model. The negative sample of the non-motorized vehicle detection model and the non-motorized vehicle detection model correspond to The false detection samples are used to train the human detection model.

It can be understood that the two conditions for detecting whether a non-motor vehicle in the non-motor vehicle area carries a person in the embodiment of the present application are “a pre-trained human detection model” and “a non-motor vehicle type in the non-motor vehicle area”. "Category" can be used in combination to make it clearer that the test result belongs to one of no human, normal human and abnormal human.

It can be seen that this solution first identifies the non-motorized vehicle area and then identifies whether the non-motorized vehicle carries people or even abnormally carried people in the non-motorized vehicle area, so that the identification of the carried people is well targeted, so it can be detected quickly and efficiently Whether non-motor vehicles carry people or even abnormally.

Corresponding to the foregoing method embodiments, the embodiments of the present application further provide a non-motorized vehicle-mounted person detection device. As shown in FIG. 4, a non-motorized vehicle-mounted person detection device provided in an embodiment of the present application may include: an image obtaining unit 410, a non-motor vehicle area obtaining unit 420, and a manned detection unit 430. The image obtaining unit 410 is configured to obtain a target image to be detected. The non-motor vehicle region obtaining unit 420 is configured to perform non-motor vehicle detection on the target image to obtain a non-motor vehicle region in the target image, where the non-motor vehicle region includes information of a non-motor vehicle. The human-carrying detection unit 430 is configured to detect whether a non-motor vehicle in the non-motor vehicle area carries a person, and obtain a detection result of the non-motor vehicle area.

The device provided in the embodiment of the present application performs non-motor vehicle detection on the obtained target image to obtain a non-motor vehicle area in the target image; detects whether a non-motor vehicle in the non-motor vehicle area carries a person, and obtains the non-motor vehicle. Detection results corresponding to the motor vehicle area. It can be seen that this solution first identifies the non-motorized vehicle area and then identifies whether the non-motorized vehicle carries people in the non-motorized vehicle area, so that the identification of the human carrying is well targeted, and therefore, it can quickly and effectively detect whether the non-motorized vehicle carries people.

Optionally, the non-motor vehicle area obtaining unit 420 is specifically configured to determine an effective detection area from the target image based on preset area calibration information; perform non-motor vehicle detection on the effective detection area to obtain the The non-motor vehicle area is described.

Optionally, the non-motor vehicle area obtaining unit 420 is specifically configured to perform non-motor vehicle detection on the target image based on a pre-trained non-motor vehicle detection model to obtain the non-motor vehicle area.

Optionally, the human-carrying detection unit 430 is specifically configured to perform area expansion processing on the non-motor vehicle area to obtain an expanded area; detect whether a non-motor vehicle in the expanded area is carrying a person, and obtain Test results in the motor vehicle area.

Optionally, the human-carrying detection unit 430 is specifically configured to detect whether a non-motor vehicle in the non-motor vehicle area is carrying a person based on a pre-trained human-carrying detection model, and obtain detection of the non-motor vehicle area. result.

Optionally, the human-carrying detection unit 430 shown is specifically: determining a human-carrying confidence level of the non-motor vehicle in the non-motor vehicle area based on a pre-trained human-carrying detection model; when the human-carrying confidence level is greater than When the confidence threshold is set, it is determined that the detection result of the non-motor vehicle area is a person.

Optionally, the shown manned detection unit 430 is specifically configured to determine whether the number of people in the non-motor vehicle area is greater than one when the manned confidence level is not greater than a preset confidence threshold, and if it is , Determining that the detection result of the non-motor vehicle area is a manned person.

Optionally, the shown human-carrying detection unit 430 is specifically configured to: determine whether the number of areas of people in the non-motorized vehicle area is greater than 1; if the number of areas of people in the non-motorized vehicle area is greater than 1, determine whether The detection result of the non-motor vehicle area is a person.

Optionally, the non-motor vehicle region obtaining unit 420 is specifically configured to perform non-motor vehicle detection on the target image based on a pre-trained non-motor vehicle detection model to obtain a premise of the non-motor vehicle region in the target image. Next, the device provided in the embodiment of the present application may further include a vehicle type analysis unit 440. The vehicle type analysis unit 440 is configured to obtain a vehicle type category of the non-motor vehicle in the non-motor vehicle area; and when the vehicle type category does not conform to a preset category, the manned detection unit 430 is triggered. The preset category includes at least one manned category

Optionally, the apparatus provided in the embodiment of the present application may further include a labeling unit 450. The labeling unit 450 is configured to label the detection result corresponding to the non-motor vehicle area and the non-motor vehicle area in the target image.

Corresponding to the foregoing method embodiments, an embodiment of the present application further provides an electronic device. As shown in FIG. 5, the electronic device includes an internal bus 510, a memory 520, a processor 530, and a communication interface ( Communications Interface) 540. The processor 530, the communication interface 540, and the memory 520 complete communication with each other through the internal bus 510.

The memory 520 is configured to store a machine feasible instruction corresponding to a non-motorized vehicle detection method.

The processor 530 is configured to read the machine-readable instructions on the memory 520 and execute the instructions to implement a non-motor vehicle-mounted person detection method provided in the present application. A non-motor vehicle-mounted person detection method may include: obtaining a target image to be detected; performing non-motor vehicle detection on the target image to obtain a non-motor vehicle region in the target image; wherein the non-motor vehicle region The motor vehicle area contains information of a non-motor vehicle; detecting whether a non-motor vehicle in the non-motor vehicle area carries a person, and obtaining a detection result of the non-motor vehicle area.

In this embodiment, for the related description of the specific steps of the non-motorized vehicle-borne person detection method, reference may be made to the description in the method embodiment provided in this application, and details are not described herein.

For details about the implementation process of the functions and functions of the units in the above device, refer to the implementation process of the corresponding steps in the foregoing method for details, and details are not described herein again.

As for the device embodiment, since it basically corresponds to the method embodiment, the relevant part may refer to the description of the method embodiment. The device embodiments described above are only schematic, and the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located One place, or it can be distributed across multiple network elements. Some or all of these modules can be selected according to actual needs to achieve the purpose of the solution of this application. Those of ordinary skill in the art can understand and implement without creative efforts.

An embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the program is processed by a processor, the method for detecting a non-motor vehicle-mounted person described in the foregoing method embodiment is implemented.

The above are only preferred embodiments of this application, and are not intended to limit this application. Any modification, equivalent replacement, or improvement made within the spirit and principle of this application shall be included in this application Within the scope of protection.

Claims

A non-motorized vehicle-borne person detection method includes:

Obtaining a target image to be detected;

Performing non-motor vehicle detection on the target image to obtain a non-motor vehicle area in the target image; wherein the non-motor vehicle area contains information of a non-motor vehicle;

Detecting whether a non-motor vehicle in the non-motor vehicle area carries a person, and obtaining a detection result of the non-motor vehicle area.
The method according to claim 1, wherein performing non-motor vehicle detection on the target image to obtain the non-motor vehicle area in the target image comprises:

Determining a valid detection area from the target image based on preset area calibration information;

Non-motor vehicle detection is performed on the effective detection area to obtain the non-motor vehicle area.
The method according to claim 1, wherein performing non-motor vehicle detection on the target image to obtain the non-motor vehicle area in the target image comprises:

Based on the pre-trained non-motor vehicle detection model, non-motor vehicle detection is performed on the target image to obtain the non-motor vehicle area.
The method according to claim 1, wherein performing non-motor vehicle detection on the target image to obtain the non-motor vehicle area in the target image comprises:

Determining a valid detection area from the target image based on preset area calibration information;

Based on the pre-trained non-motor vehicle detection model, non-motor vehicle detection is performed on the effective detection area to obtain the non-motor vehicle area.
The method according to any one of claims 1-4, wherein detecting whether a non-motor vehicle in the non-motor vehicle area carries a person and obtaining a detection result of the non-motor vehicle area comprises:

Performing area expansion processing on the non-motor vehicle area to obtain an expanded area;

It is detected whether the non-motor vehicle in the expanded area is carrying a person, and a detection result of the non-motor vehicle area is obtained.
The method according to any one of claims 1-4, wherein detecting whether a non-motor vehicle in the non-motor vehicle area carries a person and obtaining a detection result of the non-motor vehicle area comprises:

Based on a pre-trained human-carrying detection model, it is detected whether a non-motorized vehicle in the non-motorized vehicle area carries a person, and a detection result of the non-motorized vehicle area is obtained.
The method according to claim 6, characterized in that, based on the pre-trained human-carrying detection model, it is detected whether a non-motor vehicle in the non-motor vehicle area is carrying a person, and the non-motor vehicle area is detected. The results include:

Determining a manned confidence level of the non-motorized vehicle in the non-motorized vehicle region based on a pre-trained human detection model;

When the manned confidence level is greater than a preset confidence threshold, it is determined that the detection result of the non-motorized area is a manned condition.
The method according to claim 7, characterized in that, based on the pre-trained human-carrying detection model, it is detected whether a non-motor vehicle in the non-motor vehicle area is carrying a person, and the non-motor vehicle area is detected. The results also include:

When the manned confidence is not greater than the confidence threshold, determining whether the number of people in the non-motor vehicle area is greater than one,

If yes, it is determined that the detection result of the non-motor vehicle area is manned.
The method according to claim 6, characterized in that, based on the pre-trained human-carrying detection model, it is detected whether a non-motor vehicle in the non-motor vehicle area is carrying a person, and the non-motor vehicle area is detected. The results also include:

Determining whether the number of people's areas in the non-motor vehicle area is greater than one;

If the number of people's areas in the non-motor vehicle area is greater than 1, determining that the detection result of the non-motor vehicle area is manned.
The method according to claim 1, further comprising:

Obtaining the type of the non-motor vehicle in the non-motor vehicle area;

When the type of the vehicle model does not meet the preset category, a step of detecting whether a non-motor vehicle in the non-motor vehicle area carries a person is performed.
The method according to any one of claims 1-4, further comprising:

Annotate the non-motor vehicle area and a detection result of the non-motor vehicle area in the target image.
A non-motorized vehicle-mounted person detection device includes:

An image obtaining unit, configured to obtain a target image to be detected;

A non-motor vehicle region obtaining unit is configured to perform non-motor vehicle detection on the target image to obtain a non-motor vehicle region in the target image; wherein the non-motor vehicle region includes information of a non-motor vehicle;

The human-carrying detection unit is configured to detect whether a non-motor vehicle in the non-motor vehicle area carries a person, and obtain a detection result of the non-motor vehicle area.
The device according to claim 12, wherein the non-motor vehicle area obtaining unit is specifically configured to:

Based on a pre-trained non-motor vehicle detection model, non-motor vehicle detection is performed on the target image to obtain a non-motor vehicle region in the target image.
The device according to claim 12, wherein the human carrying detection unit is specifically configured to:

Based on a pre-trained human-carrying detection model, it is detected whether a non-motorized vehicle in the non-motorized vehicle area carries a person, and a detection result of the non-motorized vehicle area is obtained.
The device according to claim 12, wherein the human carrying detection unit is specifically configured to:

Perform area expansion processing on the non-motor vehicle area to obtain an expanded area; detect whether a non-motor vehicle in the expanded area is carrying a person, and obtain a detection result of the non-motor vehicle area.
The device according to any one of claims 12-15, further comprising:

The labeling unit is configured to label the non-motor vehicle area and the detection result corresponding to the non-motor vehicle area in the target image.
An electronic device includes: an internal bus, a memory, a processor, and a communication interface; wherein the processor, the communication interface, and the memory communicate with each other through the internal bus; wherein the memory, It is used to store the machine feasible instructions corresponding to the non-motorized vehicle detection method;

The processor is configured to read the machine-readable instructions on the memory and execute the instructions to implement the non-motor vehicle-mounted person detection method according to any one of claims 1-11.
A computer-readable storage medium stores a computer program thereon, and when the computer program is processed by a processor, the non-motor vehicle-mounted person detection method according to any one of claims 1-11 is implemented.