CN113537183B - Electric vehicle helmet identification method based on artificial intelligence - Google Patents
Electric vehicle helmet identification method based on artificial intelligence Download PDFInfo
- Publication number
- CN113537183B CN113537183B CN202111094693.1A CN202111094693A CN113537183B CN 113537183 B CN113537183 B CN 113537183B CN 202111094693 A CN202111094693 A CN 202111094693A CN 113537183 B CN113537183 B CN 113537183B
- Authority
- CN
- China
- Prior art keywords
- points
- human body
- helmet
- straight line
- electric vehicle
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 23
- 238000001514 detection method Methods 0.000 claims abstract description 21
- 238000012549 training Methods 0.000 claims description 7
- 210000003127 knee Anatomy 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 5
- 238000002372 labelling Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 210000000689 upper leg Anatomy 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to an electric vehicle helmet identification method based on artificial intelligence, which is characterized in that a current road scene graph is acquired through a camera, key nodes of a human body in a road scene are estimated and identified through a human body posture technology, a coordinate system is established, region division is carried out according to key node coordinates of the head, a fitting straight line of the head coordinates of the human body is obtained by using a least square method, a co-multiplication key node set and a non-co-multiplication key node set are obtained according to the fitting straight line, and whether electric vehicle personnel wear a helmet according to the specification is identified and judged by combining a target detection network.
Description
Technical Field
The application relates to the field of artificial intelligence, in particular to an electric vehicle helmet identification method based on artificial intelligence.
Background
Whether the detection of wearing the helmet to electric motor car personnel at present mainly has two kinds, one kind is the artifical detection of staff such as traffic police, and the shortcoming is: limited police strength is used for checking the electric vehicle riders and passengers with huge quantity, time and labor are wasted, and the working efficiency is low; the other is to utilize a video monitoring system, acquire image information through a roadside traffic camera, utilize a DNN network to identify the condition of a driver of an electric vehicle, detect whether a driving tool is the electric vehicle or not, and whether a corresponding target wears a helmet or not in detection, wherein the defects are as follows: when the image is acquired from the rear upper side by the road camera, the phenomenon that the electric vehicle carries people due to more targets in the image cannot be judged whether the phenomenon is illegal or not.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide an electric vehicle helmet identification method based on artificial intelligence.
In order to achieve the purpose, the invention adopts the following technical scheme: an electric vehicle helmet identification method based on artificial intelligence comprises the following steps:
acquiring a scene picture of a non-motor vehicle lane;
obtaining human body key nodes in a non-motor vehicle lane scene through human body posture estimation, and establishing a coordinate system to obtain the coordinates of head key nodes in each human body key node;
performing region division according to the head key node coordinates, performing linear fitting on the head key node coordinates in each region to obtain a plurality of fitting straight lines, classifying the plurality of fitting straight lines to obtain a co-multiplication relation fitting straight line and a non-co-multiplication relation fitting straight line, and obtaining a co-multiplication key node set and a non-co-multiplication key node set through the classified co-multiplication relation fitting straight lines and non-co-multiplication relation fitting straight lines;
the method comprises the steps that whether a corresponding human body in all head key nodes wears a helmet or not is detected by using a target detection network, when the fact that the human body does not wear the helmet is detected, if the head key nodes of the human body without the helmet are located in a shared key node set, an event that the human body does not wear the helmet is violated, if the head key nodes of the human body without the helmet are located in a non-shared key node set, whether a person corresponding to the event that the human body does not wear the helmet uses a vehicle or not is detected, and if the vehicle is an electric vehicle, the event that the human body does not wear the helmet is violated.
The straight line fitting method comprises the following steps:
numbering key nodes of the head of a human body according to the size of coordinates, and initializing a connection mark of a head position point
And detecting the state
=0, selecting the point with the minimum label, and selecting the point in one row with the vertical coordinate difference not more than half of the length of the electric vehicle to divide the point into an area;
computing
The average value of the vertical coordinates of all points in the area is divided into an area by selecting the points with the difference between the vertical coordinates and the average value not more than the length of an electric vehicle, and the points are set
Selecting the points with the smallest mark number from the rest undetected points in the region
Repeating the above steps until there are no undetected points in the image;
and traversing each region in sequence, performing linear fitting on points in each region by using a least square method to obtain a fitting straight line, wherein the fitting straight line is a co-multiplication relation fitting straight line if the fitting straight line has internal points and external points, and the fitting straight line is a non-co-multiplication relation fitting straight line if the fitting straight line has internal points and no external points.
The co-multiplication key node set comprises outer points in all co-multiplication relation fitting straight lines and inner points with the shortest distance from the outer points, and the non-co-multiplication key node set comprises inner points in all non-co-multiplication relation fitting straight lines and inner points without corresponding outer points in all co-multiplication relation fitting straight lines.
The human body posture estimation is carried out through a human body key point detection network of an Encoder-Decoder structure,
and taking the trained road scene image information as input, coding and decoding, wherein labels are 13 key points of feet, knees, crotch, shoulders, elbows, hands, heads and the like of a human body, and the Loss function is obtained by adopting mean square error Loss function training.
The target detection network uses an Encoder-FC network and sets the head coordinates of the human body as
The coordinates of both shoulders are respectively
，
To do so by
Is used as the center of the device,
，
and marking and training the rectangular frame cut out from the vertex to obtain a target with a label of 0, and determining the target with the label of 1 as a non-helmet target, and determining the target with the label of 1 as a helmet target.
The vehicle detection method comprises the following steps:
obtaining the position coordinates of the head and the two hands of each suspicious object, and setting the coordinates of the head as
The position coordinates of both hands are
To do so by
，
，
，
And (4) making a rectangular frame as a vertex, detecting the rectangular frame area by using an Encoder-FC network, and if the rectangular frame area has the electric vehicle rearview mirror, indicating that the vehicle is an electric vehicle.
The invention has the beneficial effects that: based on artificial intelligence, a fitting straight line is obtained by combining a least square method through a human body posture estimation technology, a co-multiplication relation key point set and a non-co-multiplication relation key point set are obtained according to the fitting straight line, the position relation between a human head key point and the set and a vehicle are analyzed by combining a target detection network, the helmet condition is identified, whether violation occurs or not is judged, the detection cost is greatly saved, and the method is accurate and efficient.
Drawings
Fig. 1 is a schematic view of an application scenario in an electric vehicle helmet identification method based on artificial intelligence of the present invention.
FIG. 2 is a flow chart of an electric vehicle helmet identification method based on artificial intelligence.
Fig. 3 is a schematic diagram of a head mark in the artificial intelligence-based electric vehicle helmet identification method.
Fig. 4 is a schematic diagram of the coordinates of the head after the background is removed in the method for identifying the electric vehicle helmet based on artificial intelligence.
Fig. 5 is a schematic view of a head number in the method for identifying an electric vehicle helmet based on artificial intelligence according to the present invention.
FIG. 6 is a schematic diagram of a non-co-multiplication relationship fitting straight line in the method for identifying the electric vehicle helmet based on artificial intelligence.
FIG. 7 is a schematic diagram of a common multiplicative relation fitting straight line in the electric vehicle helmet identification method based on artificial intelligence.
Fig. 8 is a schematic diagram of connection of a common-riding key point in an application scenario in the method for identifying an electric vehicle helmet based on artificial intelligence of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the application, as detailed in the appended claims.
Example 1:
in the embodiment, as shown in fig. 1, the application scenario is that there are more vehicles in the non-motor lane.
The embodiment provides an electric vehicle helmet identification method based on artificial intelligence as shown in fig. 2, which includes:
acquiring a scene picture of a non-motor vehicle lane;
the method comprises the step of shooting a picture of a non-motor vehicle lane in a current scene through a camera at a high position of a road junction.
Obtaining human body key nodes in a non-motor vehicle lane scene through human body posture estimation, and establishing a coordinate system to obtain the coordinates of head key nodes in each human body key node;
the purpose of this step is to obtain the human key point position in the picture, facilitate the subsequent processing.
The key point detection network of the human body is an Encoder-Decoder structure, trained road scene image information is input, encoded and decoded, and tags are 13 types of key points of the human body, such as feet, knees, thighs, shoulders, elbows, hands, heads and the like.
Wherein, the labeling process is as follows: each type of key point corresponds to a single channel, the positions of the corresponding pixels of the key points are marked in the channel, and then Gaussian blur is adopted to enable the key points to be formed at the marked points. The Loss function is trained by adopting a mean square error Loss function.
Performing region division according to the head key node coordinates, performing linear fitting on the head key node coordinates in each region to obtain a plurality of fitting straight lines, classifying the plurality of fitting straight lines to obtain a co-multiplication relation fitting straight line and a non-co-multiplication relation fitting straight line, and obtaining a co-multiplication key node set and a non-co-multiplication key node set through the classified co-multiplication relation fitting straight lines and non-co-multiplication relation fitting straight lines;
the purpose of this step is to detect the set of points where there is an electrical manned relationship and points where there is no electrical manned relationship among all the points.
The straight line fitting method comprises the following steps:
(1) initializing data
Marking the human head detected by the key point with the lower left corner of the image as the origin as shown in figure 3, removing the background to obtain figure 4, marking the human head detected by the road from left to right and from bottom to top as shown in figure 5, wherein the marking rule is that the longitudinal coordinates of each point are sorted, and for the points with the same longitudinal coordinate, the points with the same longitudinal coordinate are sorted, and then the points with the same longitudinal coordinate are markedThe abscissa thereof is sorted, and the corresponding coordinate is
，
，
，……,
Each point is provided with a mark representing connection relation
And a flag indicating the detection state
The initial value is 0, which means that the point is not connected and is not detected.
(2) The method comprises the following steps of dividing the regions in rows, wherein in a straight road scene, the distance between the heads of electric vehicle drivers at the front position and the rear position can be approximate to consider that the distance between the heads of the electric vehicle drivers at the front position and the rear position is different by the length of an electric vehicle body, each row of all electric vehicles can be divided into one region, and the specific division steps are as follows:
a
the point of the whole area with the smallest label, i.e. the point with the label 1 in fig. 5, is taken
The difference between the vertical coordinates of the vehicles in a row cannot exceed half the vehicle length, otherwise the accuracy of straight line fitting cannot be ensured, and the vertical coordinate is selected to be positioned
Points within the region, divided into a group, wherein
According to the camera imaging principle, the real scene is
The length of the vehicle body is projected to the length of the vehicle body at the corresponding position in the image at different distances from the camera;
c. Selecting one point with the smallest label from the rest undetected points, i.e. the detection state mark
And repeating the above steps again until no undetected point exists in the image, that is, no point with the minimum mark number value exists
A point of (a);
it should be noted that the length of the vehicle body of the electric vehicle commonly found on the market is within
Between, the bicycle body length is
However, the electric vehicle of people is gradually and mostly small, so that the general body length of the bicycle or the electric vehicle in the scene can be set as
According to the camera imaging principle, the length of the image corresponding to different positions is obtained as a threshold value
。
(3) Traversing each region in sequence, and performing straight line fitting on points in the region by using a least square method, wherein the fitting principle is as follows:
wherein,
is a pre-selected set of linearly independent functions,
is the coefficient to be determined and is,
the fitting criterion is that
And
is a distance of
The sum of squares of (a) is the smallest, where the points that are finally selected for straight line fitting are interior points and the points that are not selected for the straight line are exterior points.
And further classifying the fitted straight line, wherein the fitted straight line is a co-multiplication relation fitted straight line if the fitted straight line has an inner point and an outer point, and the fitted straight line is a non-co-multiplication relation fitted straight line if the fitted straight line has an inner point and no outer point.
The common-multiplication key points are all inner points on the fitting straight line in a traversal way for one point outside the fitting straight line, a connection relation is established between one inner point with the minimum distance and the outer point, at the moment, the common-multiplication relation exists between the inner point and the outer point, connection marks are mutually arranged, the inner points and the outer point are the common-multiplication key points, the non-common-multiplication key points are included in the non-common-multiplication relation fitting straight line, the inner points on the fitting straight line and the inner points without corresponding outer points in the common-multiplication relation fitting straight line can obtain a common-multiplication key point set and a non-common-multiplication key point set.
Further exemplifying the use of co-multiplying relationship fitting straight line and non-co-multiplying relationship fitting straight line and co-multiplying key points, for example, fitting the points in the first area in fig. 5 to obtain fig. 6, from which it can be seen that points 1, 2, and 3 are inner points, and the difference in ordinate is 0.5
There are no outer points in the range, the fitted straight line is considered as a fitted straight line with a non-co-multiplication relation, the point 1, the point 2 and the point 3 are non-co-multiplication key points, then one point 4 with the minimum label is selected from the remaining undetected points, the dividing step is the same as the previous step, as shown in fig. 7, except that after the group of points is divided, the point
Point of contact
Point of contact
Point of contact
And if an outer point appears for fitting the inner point of the straight line, and the point 8 is the point, the fitted straight line is regarded as a co-multiplication relation fitted straight line. At that time, calculate
The Euclidean distance from the number point to other points can be obtained
Number is clicked to
The distance between the signal points is minimal, so it is considered that
Number point and
point No. 8 and point No. 6 actually multiply one electric vehicle together as shown in fig. 8 and indicated by an orange line in comparison with the original image 3, in the two regions of fig. 6 and 7, the shared key points are set as point 8 and point 6 in fig. 7, and the non-shared key points are set as point 1, point 2, point 3 in fig. 6 and point 4, point 5 and point 7 in fig. 7.
The method comprises the steps that whether a corresponding human body in all head key nodes wears a helmet or not is detected by using a target detection network, when the fact that the human body does not wear the helmet is detected, if the head key nodes of the human body without the helmet are located in a shared key node set, an event that the human body does not wear the helmet is violated, if the head key nodes of the human body without the helmet are located in a non-shared key node set, whether a vehicle is used or not is detected for a person corresponding to the event that the human body does not wear the helmet, and if the vehicle is an electric vehicle, the event that the human body does not wear the helmet is violated.
The step aims to judge whether the target without the helmet violates the rules or not, because in the practical situation, most of the images shot by the roadside cameras block the lower part of the vehicle, the blocking of the head of the upper part and the rearview mirrors of the vehicle is small, the suspicious object can be further identified whether the rearview mirrors exist in the driving vehicle or not, the area corresponding to the human body is the area from shoulders to hands, and the situation that the rearview mirrors of the electric vehicle exist in the area is only detected.
The target network detection method comprises the following steps:
(1) let the head coordinate of the human body be
The coordinates of both shoulders are respectively
，
(ii) a To be provided with
Is used as the center of the device,
，
making a rectangular frame for the vertex;
（
2) an Encoder-FC network is used, a curve with the width of 1 pixel is marked on a rectangular frame to serve as a label of a corresponding helmet, namely the center point of helmet pixels in each row of pixels serves as a label and is marked with 1, the other lines of pixels serve as 0,
the Loss function of the cross entropy is adopted for training;
(3) the target labeled 1 after training is considered a helmet-worn target, and the target labeled 0 is considered a helmet-free target.
The method for judging whether the violation exists is as follows:
detecting helmet-less target connection indicia
If, if
If the target head key node is located in the shared key node set, the electric vehicle carries people and no helmet is carried by the people, and violation is judged; if it is
If the target head key node is located in the non-shared key node set, the vehicle needs to be further detected, and if the vehicle is an electric vehicle, violation exists, otherwise, the violation does not exist.
Further illustrated, the method of detecting a vehicle is as follows:
(1) acquiring the head of each suspicious object and the position coordinates of both hands, and setting the coordinates of the head as
The position coordinates of both hands are
To do so by
，
，
，
And (4) making a rectangular frame as a vertex, and detecting whether the rearview mirror of the electric vehicle exists in the area.
(2) And (3) labeling the rectangular frame after the first step of processing by using an Encoder-FC network, and further labeling a curve with the width of 1 pixel as a label corresponding to the electric vehicle rearview mirror, namely, the central point of the electric vehicle rearview mirror pixel in each row of pixels is a label and is labeled as 1, the other pixels are labeled as 0, and the Loss function adopts a cross entropy Loss function.
For further example, as shown in fig. 8, it is detected that the helmet is not worn at the point 6 and the point 8, and the critical head nodes of the two are located in the riding-sharing critical node set, it is determined that there is a violation, it is detected that the helmet is not worn at the point 3, and the critical head nodes of the point 3 are located in the non-riding-sharing critical node set, and it is further detected that the vehicle at the point 3 is an electric vehicle, and it is determined that there is a violation.
The second embodiment is provided based on the first embodiment of the invention, and comprises an information acquisition unit, a driving unit and a driving unit, wherein the information acquisition unit acquires lane images through a camera erected at the high position of a non-motor lane; the information processing unit is used for sending the image information to the cloud server, analyzing and processing the image information through a program, and judging whether the violation phenomenon that the electric vehicle personnel do not wear the helmet exists in the image; the prompting unit is used for prompting or notifying through audio output equipment erected at a high position of a lane when a current person violates a rule, such as: the method is characterized in that a video image display device is arranged beside a road, and images of persons with violation are displayed and prompt notification is carried out by matching with voice.
In practical application, this scheme can be applied to traffic lights department crossing department in the scene of non-motor vehicle lane violation detection, to riding passerby in a large number, need not artifical detection whether for riding electric motor car object, whether not wear the helmet according to the regulation, the picture that only needs the camera to shoot is handled and is analyzed, can judge out the violation incident, based on artificial intelligence, accurate and high-efficient.
The above embodiments are merely illustrative of the present invention, and should not be construed as limiting the scope of the present invention, and all designs identical or similar to the present invention are within the scope of the present invention.
Claims (6)
1. An electric vehicle helmet identification method based on artificial intelligence is characterized by comprising the following steps:
acquiring a scene picture of a non-motor vehicle lane;
obtaining human body key nodes in a non-motor vehicle lane scene through human body posture estimation, and establishing a coordinate system to obtain the coordinates of head key nodes in each human body key node;
performing region division according to the head key node coordinates, performing linear fitting on the head key node coordinates in each region to obtain a plurality of fitting straight lines, classifying the plurality of fitting straight lines to obtain a co-multiplication relation fitting straight line and a non-co-multiplication relation fitting straight line, and obtaining a co-multiplication key node set and a non-co-multiplication key node set through the classified co-multiplication relation fitting straight lines and non-co-multiplication relation fitting straight lines;
the method comprises the steps that whether a corresponding human body in all head key nodes wears a helmet or not is detected by using a target detection network, when the fact that the human body does not wear the helmet is detected, if the head key nodes of the human body without wearing the helmet are located in a shared key node set, an event that the human body does not wear the helmet is violated, if the head key nodes of the human body without wearing the helmet are located in a non-shared key node set, whether a person corresponding to the event that the human body does not wear the helmet uses a vehicle or not is detected, and if the vehicle is an electric vehicle, the event that the human body does not wear the helmet is violated.
2. The artificial intelligence-based electric vehicle helmet identification method according to claim 1, wherein the straight line fitting method comprises:
numbering key nodes of the head of a human body according to the size of coordinates, and initializing a connection mark of a head position point
And detecting the state
=0, selecting the point with the minimum label, and selecting the point in one row with the vertical coordinate difference not more than half of the length of the electric vehicle to divide the point into an area;
computing
The average value of the vertical coordinates of all points in the area is selected, and the difference between the vertical coordinates and the average value is not more than one electric vehicleA point of the vehicle length, divided into a region, set to
Selecting the points with the smallest mark number from the rest undetected points in the region
Repeating the above steps until there are no undetected points in the image;
and traversing each region in sequence, performing linear fitting on points in each region by using a least square method to obtain a fitting straight line, wherein the fitting straight line is a co-multiplication relation fitting straight line if the fitting straight line has internal points and external points, and the fitting straight line is a non-co-multiplication relation fitting straight line if the fitting straight line has internal points and no external points.
3. The artificial intelligence-based electric vehicle helmet identification method according to claim 1, wherein the co-multiplication key node set comprises outer points in all co-multiplication relation fitting straight lines and inner points closest to the outer points, and the non-co-multiplication key node set comprises inner points in all non-co-multiplication relation fitting straight lines and inner points without corresponding outer points in all co-multiplication relation fitting straight lines.
4. The method as claimed in claim 1, wherein the human body posture estimation is performed by using a human body key point detection network of an Encoder-Decoder structure, using image information of a road scene after training as input, encoding and decoding, wherein labels are 13 types of key points of feet, knees, crotch, shoulders, elbows, hands and head of a human body, and a Loss of mean square function (Loss of mean square) function is used for training to obtain the Loss of human body posture estimation.
5. The artificial intelligence based electric vehicle helmet identification method according to claim 1, wherein the target detection network is an Encoder-FC network, and human head coordinates are set as
The coordinates of both shoulders are respectively
，
To do so by
Is used as the center of the device,
，
and marking and training the rectangular frame cut out from the vertex to obtain a target with a label of 0, and determining the target with the label of 1 as a non-helmet target, and determining the target with the label of 1 as a helmet target.
6. The artificial intelligence based electric vehicle helmet identification method according to claim 1, wherein the vehicle detection method comprises the following steps:
obtaining the position coordinates of the head and the two hands of each suspicious object, and setting the coordinates of the head as
The position coordinates of both hands are
To do so by
，
，
，
And (4) making a rectangular frame as a vertex, detecting the rectangular frame area by using an Encoder-FC network, and if the rectangular frame area has the electric vehicle rearview mirror, indicating that the vehicle is an electric vehicle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111094693.1A CN113537183B (en) | 2021-09-17 | 2021-09-17 | Electric vehicle helmet identification method based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111094693.1A CN113537183B (en) | 2021-09-17 | 2021-09-17 | Electric vehicle helmet identification method based on artificial intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113537183A CN113537183A (en) | 2021-10-22 |
| CN113537183B true CN113537183B (en) | 2021-12-10 |
Family
ID=78092828
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111094693.1A Active CN113537183B (en) | 2021-09-17 | 2021-09-17 | Electric vehicle helmet identification method based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113537183B (en) |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105760847B (en) * | 2016-03-01 | 2019-04-02 | 江苏大学 | A kind of visible detection method of pair of helmet of motorcycle driver wear condition |
| CN111310653A (en) * | 2020-02-13 | 2020-06-19 | 上海眼控科技股份有限公司 | Detection method and device for wearing helmet, computer equipment and storage medium |
| CN112329532A (en) * | 2020-09-30 | 2021-02-05 | 浙江汉德瑞智能科技有限公司 | Automatic tracking safety helmet monitoring method based on YOLOv4 |
| CN113160575A (en) * | 2021-03-15 | 2021-07-23 | 超级视线科技有限公司 | Traffic violation detection method and system for non-motor vehicles and drivers |
-
2021
- 2021-09-17 CN CN202111094693.1A patent/CN113537183B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113537183A (en) | 2021-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Borkar et al. | A novel lane detection system with efficient ground truth generation | |
| CN111179345B (en) | Front vehicle line-crossing violation behavior automatic detection method and system based on vehicle-mounted machine vision | |
| CN108062349B (en) | Video monitoring method and system based on video structured data and deep learning | |
| CN109389046B (en) | All-weather object identification and lane line detection method for automatic driving | |
| Siogkas et al. | Traffic lights detection in adverse conditions using color, symmetry and spatiotemporal information | |
| EP3751480B1 (en) | System and method for detecting on-street parking violations | |
| CN108596054A (en) | A kind of people counting method based on multiple dimensioned full convolutional network Fusion Features | |
| CN110135377B (en) | Method and device for detecting motion state of object in vehicle-road cooperation and server | |
| CN111797803A (en) | Road guardrail abnormity detection method based on artificial intelligence and image processing | |
| CN105184301B (en) | A kind of method that vehicle heading is differentiated using four-axle aircraft | |
| CN113658427A (en) | Road condition monitoring method, system and equipment based on vision and radar | |
| Supreeth et al. | An approach towards efficient detection and recognition of traffic signs in videos using neural networks | |
| CN107918775B (en) | Zebra crossing detection method and system for assisting safe driving of vehicle | |
| Balali et al. | Video-based detection and classification of US traffic signs and mile markers using color candidate extraction and feature-based recognition | |
| JP2013089129A (en) | Mobile object detection device, computer program, and mobile object detection method | |
| TWI851992B (en) | Object tracking integration method and integrating apparatus | |
| CN113537183B (en) | Electric vehicle helmet identification method based on artificial intelligence | |
| Suttiponpisarn et al. | Detection of wrong direction vehicles on two-way traffic | |
| CN114724122A (en) | Target tracking method and device, electronic equipment and storage medium | |
| CN113963373A (en) | Video image dynamic detection and tracking algorithm based system and method | |
| Badura et al. | Intelligent traffic system: Cooperation of MANET and image processing | |
| CN112528944A (en) | Image identification method and device, electronic equipment and storage medium | |
| CN117612117A (en) | Roadside near weed segmentation method, system and medium based on vehicle-mounted recorder | |
| CN107901907B (en) | Lane line detection method based on color block statistics | |
| CN106462762B (en) | Based on vehicle detection, tracking and the positioning for enhancing anti-perspective transform |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |