CN113111876A - Method and system for obtaining evidence of traffic violation - Google Patents

Method and system for obtaining evidence of traffic violation Download PDF

Info

Publication number
CN113111876A
CN113111876A CN202110408305.6A CN202110408305A CN113111876A CN 113111876 A CN113111876 A CN 113111876A CN 202110408305 A CN202110408305 A CN 202110408305A CN 113111876 A CN113111876 A CN 113111876A
Authority
CN
China
Prior art keywords
violation
target object
motor vehicle
license plate
traffic
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.)
Pending
Application number
CN202110408305.6A
Other languages
Chinese (zh)
Inventor
陈磊
黄金叶
陈予涵
陈予琦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Qiyang Special Equipment Technology Engineering Co ltd
Original Assignee
Shenzhen Qiyang Special Equipment Technology Engineering Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shenzhen Qiyang Special Equipment Technology Engineering Co ltd filed Critical Shenzhen Qiyang Special Equipment Technology Engineering Co ltd
Priority to CN202110408305.6A priority Critical patent/CN113111876A/en
Publication of CN113111876A publication Critical patent/CN113111876A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
    • G06V20/63—Scene text, e.g. street names
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G06V20/54—Surveillance or monitoring of activities, e.g. for recognising suspicious objects of traffic, e.g. cars on the road, trains or boats
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/20—Movements or behaviour, e.g. gesture recognition
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/017—Detecting movement of traffic to be counted or controlled identifying vehicles
    • G08G1/0175—Detecting movement of traffic to be counted or controlled identifying vehicles by photographing vehicles, e.g. when violating traffic rules
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/04—Detecting movement of traffic to be counted or controlled using optical or ultrasonic detectors
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
    • G06V20/625—License plates
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/08—Detecting or categorising vehicles

Abstract

The invention relates to the technical field of intelligent traffic, and discloses a method and a system for obtaining evidence of traffic violation, which comprise the following steps: acquiring a road image in real time; detecting the category of a target object and the coordinates of the target object through a road image; calculating to obtain the moving track of the target object through more than 2 road images; carrying out violation behavior detection so as to detect whether a violation behavior exists in the target object; and if the violation behaviors of the target object are detected, identifying the license plate information of the violation target object, storing the violation information and the license plate information of the violation target object, and uploading the violation information and the license plate information to the cloud server. The method can be used for carrying out violation detection in real time, can work continuously for 24 hours, realizes multiple-aspect traffic violation evidence obtaining, and saves a large amount of labor cost.

Description

Method and system for obtaining evidence of traffic violation
Technical Field
The invention belongs to the technical field of intelligent traffic, and particularly relates to a method and a system for obtaining evidence of traffic violation.
Background
The state vigorously promotes the construction of smart cities, encourages the utilization of various information technologies or innovative concepts, and makes the urban systems and services open and integrated to improve the efficiency of resource application, optimize urban management and services and improve the quality of life of citizens, and the smart traffic system is the core part of the construction of the smart cities.
However, in the existing traffic system, generally, a radar detection device is used to obtain environmental information, for example, in the patent application with the application number of CN201810238011.1, a trunk cooperative annunciator control method based on a radar detector is disclosed, and a radar vehicle detector is used to implement optimization processing of a control scheme of traffic signals at multiple intersections of a trunk, where the scheme includes: installing radar detectors at each intersection; acquiring and communicating data of a single intersection; a plurality of intersection data are collected and stored in a network; the main trunk cooperates with the signal machine to control algorithm processing; signal instruction issuing service and signal lamp control.
However, radar detection equipment can only obtain the state and space coordinate information of object motion, cannot obtain the visual characteristics of an object, cannot identify a license plate number and the like, an existing traffic system generally adopts cloud computing to perform data processing, but the traffic scene has high requirement on timeliness, needs to respond to emergency in real time, cloud computing naturally has time delay, is easily influenced by network fluctuation, is not high in stability, and is difficult to perform real-time detection of violation behaviors; the conventional traffic violation detection needs a traffic police to detect the violation of the non-motor vehicle on site, and the traffic police drops into a specific road section, so that the labor cost is high, and the duration is short.
Therefore, a technical scheme is desired to solve the technical problems of difficulty in real-time detection of violation behaviors, high labor cost and short duration in the prior art.
Disclosure of Invention
In order to solve the problems, the invention aims to provide a method and a system for obtaining evidence of traffic violation, which can detect the violation behaviors in real time, can work continuously for 24 hours, realize multiple-aspect traffic violation evidence obtaining and save a large amount of labor cost.
In order to achieve the above object, the technical solution of the present invention is as follows.
A method for traffic violation forensics, comprising:
s1, acquiring a road image in real time;
s2, detecting the category of 1 or more target objects and the coordinates of each target object from the road image;
s3, calculating the moving track of each target object through more than 2 road images;
s4, carrying out violation behavior detection, thereby detecting whether a target object has violation behaviors;
and S5, if the target object is detected to have the violation behaviors, recognizing the license plate information of the violation target object, storing the violation information and the license plate information of the violation target object, and uploading the violation information and the license plate information to the cloud server.
According to the traffic violation evidence obtaining method, violation detection can be carried out in real time, after the fact that the target object has violation behaviors is detected, the license plate number of the violation target object can be identified, violation information and license plate information of the violation target object are stored and uploaded to the cloud server, uninterrupted work can be carried out for 24 hours, traffic violation evidence obtaining in various aspects is achieved, and a large amount of labor cost is saved.
Further, in S2, the categories of 1 or more target objects and the coordinates of each target object are obtained by an object detection algorithm; the object detection algorithm includes:
the object detection algorithm includes: acquiring a road image at the time t from a road end camera as an INPUT image INPUT A, inputting the INPUT image into a detectable violation target object model, and detecting to obtain a target object O1Coordinates (x) of center point of bounding box in INPUT image INPUT A1,y1) Width w of the bounding box1Height h of bounding box1And class c of object1；
The model for detecting the violation target object is obtained through the following steps:
collecting relevant pictures of traffic violation, manually marking the pictures and making a traffic violation data set;
and training a DetectNet network by using the data set to obtain a detectable violation target object model.
Further, the target object is classified into any one of a motor vehicle, a normal manned non-motor vehicle, an illegal manned non-motor vehicle, a non-motor vehicle with a normal helmet worn by a driver, a non-motor vehicle with a helmet not worn by a driver, a red signal lamp, a yellow signal lamp and a green signal lamp.
Further, in S3, the movement trajectory of the target object is obtained through an object tracking algorithm;
the object tracking algorithm includes:
acquiring two road images at t and t +1 moments from a road end camera as INPUT images INPUT A and INPUT B;
inputting images INPUT A and INPUT B by using target tracking SORT algorithm, and detecting the obtained target object O according to object detection algorithm1Coordinates (x) of center point of bounding box in INPUT image INPUT A1,y1) Width w of the bounding box1And height h of bounding box1The object O in the INPUT image INPUT B can be obtained1Corresponding target object O2Coordinate (x) of center point of bounding box2,y2) Width w of the bounding box2And height h of bounding box2And calculating to obtain the moving track of the target object.
In the invention, the object tracking algorithm is responsible for the relation between target objects in different images in continuous time.
Further, the method for traffic violation forensics further comprises the following steps:
s0, recording the established traffic rules;
and S4, judging according to the category and/or the moving track of each target object and a preset traffic rule input in advance, detecting whether the target object has a violation behavior, if so, identifying license plate information of the violation target object, processing and storing the violation information and the license plate information of the violation target object, and uploading the processed structured data to a cloud server.
Furthermore, the violation behaviors comprise non-motor vehicle retrograde motion, non-motor vehicle violation people carrying people, non-motor vehicle drivers do not wear helmets, non-motor vehicles run red light, motor vehicles run red light and the like.
For example, detecting violations of traffic signal behavior: when the detected target object comprises a red signal lamp and a motor vehicle, the moving track of the motor vehicle is obtained through an object tracking algorithm, the moving track of the motor vehicle is combined with a set traffic rule which is input in advance to judge, whether the motor vehicle has red light running violation behaviors is detected, if yes, the red light running violation information of the motor vehicle and the license plate number of the motor vehicle are processed and stored, and then the processed structured data are uploaded to a cloud server.
Another example is to detect retrograde behavior: the method comprises the steps of recording a preset traffic rule in advance, obtaining a moving track of a motor vehicle through an object tracking algorithm when a detected target object comprises the motor vehicle, judging the moving track of the motor vehicle by combining the moving track of the motor vehicle with the preset traffic rule recorded in advance, judging the violation behavior of the motor vehicle running in the reverse direction if the moving direction of the motor vehicle is opposite to the normal running direction, and processing violation information of the motor vehicle running in the reverse direction and a license plate number of the motor vehicle.
Further, identifying license plate information of the violation target object comprises:
screenshot a license plate of a target object from a road picture to serve as an INPUT image INPUT P;
and inputting the INPUT image INPUT P into a convolutional neural network, and performing feature extraction convolution operation to obtain a feature mapping image of the license plate image.
And identifying the license plate number in the license plate image by using the full-connection layer. And (3) calculating the feature mapping chart C4 obtained in the steps through two fully-connected layers, and outputting a vector with the size N x 1(N is the length of the characters of the license plate number). The vector obtains N category values through calculation operations such as activation, inverse normalization and the like, the N category values respectively represent characters (categories) of each character position of the license plate, and all the obtained characters are connected in sequence to obtain the license plate number.
The present invention also provides a system for traffic violation forensics comprising:
the image acquisition module is used for acquiring a road image in real time;
the image acquisition module is in communication connection with the edge computing terminal, the image acquisition device acquires a real-time image and transmits the real-time image to the edge computing terminal, the category and the coordinate of each target object are detected through road images in the edge computing, the moving track of each target object is calculated through more than 2 road images, and then the violation behavior detection is carried out, so that whether the violation behavior exists in each target object is detected;
and the edge extreme terminal is in communication connection with the cloud server, and if the edge extreme terminal detects that the target object has the violation behavior, the license plate information of the violation target object is identified, the violation information and the license plate information of the violation target object are stored and uploaded to the cloud server.
The image acquisition module comprises a plurality of cameras, the cameras are used for acquiring real-time images and transmitting the real-time images to the edge computing terminal through the switch, the edge computing terminal achieves the purpose of obtaining evidence of non-motor vehicles or motor vehicle violation through the method for obtaining evidence of traffic violation, and the processed structured data about the violation vehicles are uploaded to the cloud server for use by a traffic management unit.
The invention has the beneficial effects that: compared with the prior art, the method and the device can perform violation detection in real time, can identify the license plate number of the violation target object after detecting that the violation behavior exists in the target object, can store the violation information and the license plate information of the violation target object, and can upload the violation information and the license plate information to the cloud server.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below.
Fig. 1 is a schematic structural diagram of a convolutional neural network of the method for traffic violation forensics of embodiment 1.
Fig. 2 is a schematic diagram of detecting a license plate and obtaining license plate bounding box position information in the process of identifying license plate information of a violation target object in embodiment 1.
Fig. 3 is a schematic diagram of inputting an INPUT image INPUT P into a convolutional neural network in the process of identifying license plate information of a violation target object in embodiment 1.
Fig. 4 is a schematic diagram of obtaining a license plate number in the process of identifying the license plate information of the violation target object in embodiment 1.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
Example 1:
referring to fig. 1-4, the present embodiment provides a method for traffic violation forensics, comprising:
s1, acquiring a road image in real time;
s2, detecting the category of 1 or more target objects and the coordinates of each target object from the road image;
s3, calculating the moving track of each target object through more than 2 road images;
s4, carrying out violation behavior detection, thereby detecting whether a target object has violation behaviors;
and S5, if the target object is detected to have the violation behaviors, recognizing the license plate information of the violation target object, storing the violation information and the license plate information of the violation target object, and uploading the violation information and the license plate information to the cloud server.
The method for obtaining evidence of traffic violation can be used for carrying out violation detection in real time, and after detecting that the violation behaviors exist in the target object, the license plate number of the violation target object can be identified, and violation information and license plate information of the violation target object are stored and uploaded to the cloud server.
In S2 of the present embodiment, the categories of 1 or more target objects and the coordinates of each target object are obtained by an object detection algorithm; the object detection algorithm includes:
the object detection algorithm includes: acquiring a road image at the time t from a road end camera as an INPUT image INPUT A, inputting the INPUT image into a detectable violation target object model, and detecting to obtain a target object O1Coordinates (x) of center point of bounding box in INPUT image INPUT A1,y1) Width w of the bounding box1Height h of bounding box1And class c of object1；
The model for detecting the violation target object is obtained through the following steps:
collecting relevant pictures of traffic violation, manually marking the pictures and making a traffic violation data set;
and training a DetectNet network by using the data set to obtain a detectable violation target object model. Wherein, DetecNet is the existing mature technology, and can be directly used after training, and the process is not repeated herein.
Specifically, based on the diversification in the natural environment, the following elements are considered in the collection process of the pictures in the traffic violation data set: picture in data set includes one or more target objects; secondly, the illumination intensity of the pictures in the data set is diversified; diversification of background environments of pictures in the data set; fourthly, the weather conditions in the data set are diversified; the shooting angle of the data set is diversified.
And manually labeling the pictures in the data set. The content of the labeled object types is as follows: motor vehicles, non-motor vehicles, people (single person or multiple persons) on the non-motor vehicles, drivers of the non-motor vehicles wearing or not wearing helmets, distribution persons and non-motor vehicles of a plurality of different distribution companies, license plates of the non-motor vehicles, license plates of the motor vehicles, green signal lamps, red signal lamps and yellow signal lamps. The labeled bounding box information also comprises the coordinates of the upper left corner and the lower right corner of the bounding box where each target object is located.
In this section, the detected target object categories include motor vehicles, normally manned non-motor vehicles, offending manned non-motor vehicles, non-motor vehicles with helmets normally worn by the driver, non-motor vehicles without helmets worn by the driver, license plates of offending motor vehicles or non-motor vehicles, and the like.
In the present embodiment, the category of the target object is any one of a motor vehicle, a non-motor vehicle with a normal person, a non-motor vehicle with an illegal person, a non-motor vehicle with a normal helmet worn by the driver, a non-motor vehicle with no helmet worn by the driver, a red signal lamp, a yellow signal lamp, and a green signal lamp.
In S3, the moving track of the target object is obtained through an object tracking algorithm;
the object tracking algorithm includes:
acquiring two road images at t and t +1 moments from a road end camera as INPUT images INPUT A and INPUT B;
inputting images INPUT A and INPUT B by using target tracking SORT algorithm, and detecting the obtained target object O according to object detection algorithm1Coordinates (x) of center point of bounding box in INPUT image INPUT A1,y1) Width w of the bounding box1And height h of bounding box1The object O in the INPUT image INPUT B can be obtained1Corresponding target object O2Coordinate (x) of center point of bounding box2,y2) Width w of the bounding box2And height h of bounding box2And calculating to obtain the moving track of the target object.
In the invention, the object tracking algorithm is responsible for the relation between target objects in different images in continuous time.
Specifically, at time t1Using object detection algorithm to INPUT image1Performing object detection to obtain target object
Figure BDA0003019347170000071
The bounding box data of (a):
Figure BDA0003019347170000072
at time t2,t3,…,tnUsing the method mentioned in the object tracking part and at t1Data of object bounding box detected at any moment, INPUT for image2,INPUT3,…,INPUTnAnd carrying out object tracking. Bounding box data and category data for multiple objects over this period of time are available:
Figure BDA0003019347170000081
and the moving track of the available object in the period of time:
Figure BDA0003019347170000082
and their classes
Figure BDA0003019347170000083
The method for obtaining evidence of traffic violation in the embodiment uses deepCNN to complete integral calculation, is an end-to-end algorithm, is supervised, and in the specific calculation, the feature extraction is performed on the whole image, and the detection result of the whole image is calculated at one time. Wherein each cell can detect multiple objects and the object box is also derived from the image features.
In this embodiment, the method for traffic violation forensics further includes:
s0, recording the established traffic rules;
and S4, judging according to the category and/or the moving track of each target object and a preset traffic rule input in advance, detecting whether the target object has a violation behavior, if so, identifying license plate information of the violation target object, processing and storing the violation information and the license plate information of the violation target object, and uploading the processed structured data to a cloud server.
And further combining the type and/or the moving track of the target object with the existing road rule to obtain a related traffic violation behavior result. Such as when the object completes the action from the a-zone to the B-zone, the object has performed a retrograde behavior.
In this embodiment, the violation behaviors include that the non-motor vehicle runs in the wrong direction, the non-motor vehicle violates a person, the driver of the non-motor vehicle does not wear a helmet, the non-motor vehicle runs a red light, the motor vehicle runs a red light, and the like.
For example, detecting violations of traffic signal behavior: when the detected target object comprises a red signal lamp and a motor vehicle, the moving track of the motor vehicle is obtained through an object tracking algorithm, the moving track of the motor vehicle is combined with a set traffic rule which is input in advance to judge, whether the motor vehicle has red light running violation behaviors is detected, if yes, the red light running violation information of the motor vehicle and the license plate number of the motor vehicle are processed and stored, and then the processed structured data are uploaded to a cloud server.
Namely, when the object detection algorithm detects that the traffic signal lamp is red light and the object detection algorithm and the object tracking algorithm detect that the non-motor vehicle moves in the zebra crossing area, the vehicle is judged to finish the behavior of running the red light in violation of the traffic signal.
Another example is to detect retrograde behavior: the method comprises the steps of recording a preset traffic rule in advance, obtaining a moving track of a motor vehicle through an object tracking algorithm when a detected target object comprises the motor vehicle, judging the moving track of the motor vehicle by combining the moving track of the motor vehicle with the preset traffic rule recorded in advance, judging the violation behavior of the motor vehicle running in the reverse direction if the moving direction of the motor vehicle is opposite to the normal running direction, and processing violation information of the motor vehicle running in the reverse direction and a license plate number of the motor vehicle.
When the object detection algorithm and the object tracking algorithm detect that the non-motor vehicle moves in the image, the motion trail of the non-motor vehicle is further calculated according to the detected movement. Then, the coordinates of the road section and the preset road rules (the normal driving direction of the road section) recorded in advance are used for comprehensively judging whether the road section is in the reverse driving state or not. When its trajectory is contrary to the established road rules for the area it is in, the algorithm determines that it is moving in the reverse direction.
In this embodiment, the license plate information for identifying the violation target object includes:
the target object type also comprises a license plate, and the license plate can be detected and the position information of the license plate bounding box can be obtained by using an object detection algorithm (four vertex coordinates of the bounding box: (x)p1,yp1)、(xp1,yp2)、(xp2,yp1)、(xp2,yp2). As shown in fig. 2, the license plate of the target object is captured from the road picture by using the above information, and is used as an INPUT image INPUT P, width W, and height H.
As shown in fig. 3, the INPUT image INPUT P is INPUT into a convolutional neural network, and a feature extraction convolution operation is performed to obtain a feature map of the license plate image. In the method proposed by the present invention, the contents are exemplified as shown in fig. 1. As shown in fig. 1, a single image INPUT P is INPUT, and feature maps C1, C2, C3, and C4 of the image INPUT P are obtained. The sequence of operation and data processing from front to back is: INPUT P → C1 → C2 → C3 → C4.
And identifying the license plate number in the license plate image by using the full-connection layer. And (3) calculating the feature mapping chart C4 obtained in the steps through two fully-connected layers, and outputting a vector with the size N x 1(N is the length of the characters of the license plate number). The vector is subjected to calculation operations such as activation, inverse normalization and the like to obtain N category values which respectively represent characters (categories) of each character position of the license plate, and all the obtained characters are connected in sequence to obtain the license plate number, as shown in fig. 4.
Example 2:
the present embodiment provides a system for traffic violation forensics comprising:
the image acquisition module is used for acquiring a road image in real time;
the edge computing terminal is communicated and connected with the image acquisition module, the image acquisition device acquires a real-time image and transmits the real-time image to the edge computing terminal, the category and the coordinate of each target object are detected through road images in the edge computing, the moving track of each target object is calculated through more than 2 road images, and then the violation behavior detection is carried out, so that whether the violation behavior exists in each target object is detected;
and the edge extreme terminal is in communication connection with the cloud server, and if the edge extreme terminal detects that the target object has a violation behavior, the license plate information of the violation target object is identified, and the violation information and the license plate information of the violation target object are stored and uploaded to the cloud server.
The image acquisition module comprises a plurality of cameras, the cameras are used for acquiring real-time images and transmitting the real-time images to the edge computing terminal through the switch, the edge computing terminal achieves the purpose of obtaining evidence of non-motor vehicles or motor vehicle violation through the method for obtaining evidence of traffic violation, and the processed structured data about the violation vehicles are uploaded to the cloud server for use by a traffic management unit; the edge computing terminal can perform real-time computation and response, the stability is high, the transmission speed is high, a large amount of computation and data are processed at the edge, only the needed structural data are sent to the cloud, and the remote transportation volume is greatly reduced; and the device can work continuously for 24 hours, so that traffic violation evidence obtaining is realized, and a large amount of labor cost is saved.
The method for obtaining a traffic violation evidence in embodiment 1 is implemented based on the system for obtaining a traffic violation evidence in this embodiment, and the system for obtaining a traffic violation evidence of the present invention is directed to a violation: the non-motor vehicle or the motor vehicle runs in the reverse direction, the non-motor vehicle violates the traffic, the non-motor vehicle driver does not wear a helmet, the non-motor vehicle or the motor vehicle violates the traffic signal lamp, and the like. The system can comprehensively judge whether the violation behaviors are carried out or not by detecting the vehicles, analyzing the appearance characteristics and the behavior characteristics of the vehicles, tracking and re-identifying the vehicles and combining the existing road rules. After a certain object is judged to have a violation behavior, the system can take out three pictures before, during and after the violation as evidence obtaining data, and extracts the violation license plate information and the structural information of the vehicle and the driver thereof by utilizing the algorithm in the step of identifying the license plate information of the violation target object, and stores and uploads the information.
The edge computing terminal can be in communication connection with the cloud server through the 4G/5G communication module. The 4G/5G communication module is an IC chip, so that the expansion is supported to have a 4G/5G communication function, and the 4G/5G remote communication function can support the normal work of a cloud function in the system for traffic violation evidence collection, such as system cloud updating, violation data real-time uploading, system real-time response cloud commands and the like, in a traffic road section without an existing network line or a wiring condition.
The system for obtaining the traffic violation evidence in the embodiment adopts the object detection algorithm in the embodiment 1 to detect the types of target objects such as motor vehicle violation, non-motor vehicle violation, affiliated company of the distributed electric vehicles, license plates, signal lamp signals and the like.
The system for obtaining evidence of traffic violation in the embodiment adopts the object tracking algorithm in the embodiment 1 to take charge of the relation between target objects among different images in continuous time, so as to calculate the moving track of the target objects;
the system for obtaining evidence of traffic violation in the embodiment adopts the violation detection step in the embodiment 1 to be responsible for detecting the traffic violation of the non-motor vehicle/motor vehicle;
the system for collecting the traffic violation in the embodiment adopts the step of identifying the license plate information of the violation target object in the embodiment 1 to identify the license plate number of the detected license plate picture.
The present invention is not limited to the above preferred embodiments, and any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. A method for traffic violation forensics, comprising:
s1, acquiring a road image in real time;
s2, detecting the category of 1 or more target objects and the coordinates of each target object from the road image;
s3, calculating the moving track of each target object through more than 2 road images;
s4, carrying out violation behavior detection, thereby detecting whether a target object has violation behaviors;
and S5, if the target object is detected to have the violation behaviors, recognizing the license plate information of the violation target object, storing the violation information and the license plate information of the violation target object, and uploading the violation information and the license plate information to the cloud server.
2. The method for traffic violation forensics according to claim 1, wherein in S2, the categories of more than 1 target object and the coordinates of each target object are obtained by an object detection algorithm; the object detection algorithm includes:
acquiring a road image at the time t from a road end camera as an INPUT image INPUT A, inputting the INPUT image INPUT A into a detectable violation target object model, and detecting to obtain the center point coordinates of a bounding box of the target object in the INPUT image INPUT A, the width of the bounding box, the height of the bounding box and the category of the target object;
the model for detecting the violation target object is obtained through the following steps:
collecting relevant pictures of traffic violation, manually marking the pictures and making a traffic violation data set;
and training a DetectNet network by using the data set to obtain a detectable violation target object model.
3. The method for traffic violation forensics according to claim 2, wherein the target object is in a category of any one of a motor vehicle, a normally manned non-motor vehicle, an illegally manned non-motor vehicle, a normally helmet-worn non-motor vehicle, a helmet-unweared non-motor vehicle, a red signal light, a yellow signal light, and a green signal light.
4. The method for traffic violation forensics according to claim 3, wherein in S3, the moving track of the target object is obtained by an object tracking algorithm;
the object tracking algorithm includes:
acquiring two road images at t and t +1 moments from a road end camera as INPUT images INPUT A and INPUT B;
by using a target tracking SORT algorithm, inputting images INPUT A and INPUT B, and according to the bounding box center point coordinate, the width of the bounding box and the height of the bounding box of the target object in the INPUT image INPUT A, which are detected by an object detection algorithm, the bounding box center point coordinate, the width of the bounding box and the height of the bounding box of the target object in the INPUT image INPUT B can be obtained, so that the moving track of the target object is calculated.
5. The method for traffic violation forensics of claim 4 wherein the method for traffic violation forensics further comprises:
s0, recording the established traffic rules;
and S4, judging according to the category and/or the moving track of each target object and a preset traffic rule input in advance, detecting whether the target object has a violation behavior, if so, identifying license plate information of the violation target object, processing and storing the violation information and the license plate information of the violation target object, and uploading the processed structured data to a cloud server.
6. The method for traffic violation forensics of claim 5, wherein the violations include non-motor vehicle violation, non-motor vehicle violation carrier, non-motor vehicle driver not wearing helmet, non-motor vehicle running red light, and motor vehicle running red light.
7. The method for obtaining the evidence of the traffic violation according to claim 6, characterized in that when the detected target object comprises a red signal lamp and a motor vehicle, the moving track of the motor vehicle is obtained through an object tracking algorithm, the moving track of the motor vehicle is combined with the preset traffic rules recorded in advance for judgment, so as to detect whether the motor vehicle has the violation behavior of running the red light, if so, the violation information of running the red light of the motor vehicle and the license plate number of the motor vehicle are processed and stored, and then the structured data obtained after the processing is uploaded to the cloud server.
8. The method for collecting the evidence of the traffic violation according to claim 6, wherein the predetermined traffic rule entered in advance comprises a normal driving direction, when the detected target object comprises a motor vehicle, a moving track of the motor vehicle is obtained through an object tracking algorithm, the moving track of the motor vehicle and the predetermined traffic rule entered in advance are combined for judgment, if the moving direction of the motor vehicle is opposite to the normal driving direction, the violation behavior that the motor vehicle runs in the reverse direction is judged, the violation information that the motor vehicle runs in the reverse direction and the license plate number of the motor vehicle are processed and stored, and then the structured data obtained after the processing are uploaded to a cloud server.
9. The method for traffic violation forensics of claim 6 wherein identifying license plate information for the violation target object comprises:
screenshot a license plate of a target object from a road picture to serve as an INPUT image INPUT P;
inputting an INPUT image INPUT P into a convolutional neural network, and performing feature extraction convolution operation to obtain a feature mapping image of the license plate image;
and recognizing the license plate number in the license plate image by using the full-connection layer, calculating the feature mapping graph obtained in the step through two full-connection layers, outputting a vector with the size of N x 1, wherein N is the character length of the license plate number, calculating the vector to obtain N category values which respectively represent the characters of each character position of the license plate, and connecting all the obtained characters in sequence to obtain the license plate number.
10. A system for traffic violation forensics, comprising:
the image acquisition module is used for acquiring a road image in real time;
the image acquisition module is in communication connection with the edge computing terminal, the image acquisition device acquires a real-time image and transmits the real-time image to the edge computing terminal, the category and the coordinate of each target object are detected through road images in the edge computing, the moving track of each target object is calculated through more than 2 road images, and then the violation behavior detection is carried out, so that whether the violation behavior exists in each target object is detected;
and the edge extreme terminal is in communication connection with the cloud server, and if the edge extreme terminal detects that the target object has the violation behavior, the license plate information of the violation target object is identified, the violation information and the license plate information of the violation target object are stored and uploaded to the cloud server.
CN202110408305.6A 2021-04-14 2021-04-14 Method and system for obtaining evidence of traffic violation Pending CN113111876A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110408305.6A CN113111876A (en) 2021-04-14 2021-04-14 Method and system for obtaining evidence of traffic violation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110408305.6A CN113111876A (en) 2021-04-14 2021-04-14 Method and system for obtaining evidence of traffic violation

Publications (1)

Publication Number Publication Date
CN113111876A true CN113111876A (en) 2021-07-13

Family

ID=76717556

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110408305.6A Pending CN113111876A (en) 2021-04-14 2021-04-14 Method and system for obtaining evidence of traffic violation

Country Status (1)

Country Link
CN (1) CN113111876A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114387533A (en) * 2022-01-07 2022-04-22 北京远度互联科技有限公司 Method and device for identifying road violation, electronic equipment and storage medium
CN114743382A (en) * 2022-06-13 2022-07-12 浙江大云物联科技有限公司 Vehicle violation behavior identification method and device based on intelligent lamp pole system

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107730904A (en) * 2017-06-13 2018-02-23 银江股份有限公司 Multitask vehicle driving in reverse vision detection system based on depth convolutional neural networks
CN109598943A (en) * 2018-12-30 2019-04-09 北京旷视科技有限公司 The monitoring method of vehicle violation, apparatus and system
CN110717433A (en) * 2019-09-30 2020-01-21 华中科技大学 Deep learning-based traffic violation analysis method and device
CN112102609A (en) * 2020-03-10 2020-12-18 中国科学院沈阳自动化研究所 Intelligent detection system for highway vehicle violation behaviors

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107730904A (en) * 2017-06-13 2018-02-23 银江股份有限公司 Multitask vehicle driving in reverse vision detection system based on depth convolutional neural networks
CN109598943A (en) * 2018-12-30 2019-04-09 北京旷视科技有限公司 The monitoring method of vehicle violation, apparatus and system
CN110717433A (en) * 2019-09-30 2020-01-21 华中科技大学 Deep learning-based traffic violation analysis method and device
CN112102609A (en) * 2020-03-10 2020-12-18 中国科学院沈阳自动化研究所 Intelligent detection system for highway vehicle violation behaviors

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
吉安卡洛•扎克尼（GIANCARLOZACCONE）等: "《TensorFlow深度学习》", 29 February 2020, 机械工业出版社 *
赵成强: "公路交通车辆违章行为检测与监控实现系统", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114387533A (en) * 2022-01-07 2022-04-22 北京远度互联科技有限公司 Method and device for identifying road violation, electronic equipment and storage medium
CN114743382A (en) * 2022-06-13 2022-07-12 浙江大云物联科技有限公司 Vehicle violation behavior identification method and device based on intelligent lamp pole system
CN114743382B (en) * 2022-06-13 2022-10-28 浙江大云物联科技有限公司 Vehicle violation behavior identification method and device based on intelligent lamp pole system

Similar Documents

Publication Publication Date Title
CN102724482B (en) Based on the intelligent vision sensing network moving target relay tracking system of GPS and GIS
CN103279756B (en) Vehicle detection based on integrated classifier analyzes system and determination method thereof
US10223911B2 (en) Video data and GIS mapping for traffic monitoring, event detection and change prediction
CN111800507A (en) Traffic monitoring method and traffic monitoring system
KR102122859B1 (en) Method for tracking multi target in traffic image-monitoring-system
CN102867417B (en) Taxi anti-forgery system and taxi anti-forgery method
EP3676754A1 (en) On-demand artificial intelligence and roadway stewardship system
CN103069434A (en) Multi-mode video event indexing
CN110837800A (en) Port severe weather-oriented target detection and identification method
CN201397576Y (en) Device for automatically shooting picture of the illegal turning of vehicles at crossings
KR102122850B1 (en) Solution for analysis road and recognition vehicle license plate employing deep-learning
CN113111876A (en) Method and system for obtaining evidence of traffic violation
CN108360442A (en) Intelligent snow-removing method, intelligent snow sweeper and computer readable storage medium
Muthanna et al. Smart system of a real-time pedestrian detection for smart city
US20220146277A1 (en) Architecture for map change detection in autonomous vehicles
Matsuda et al. A system for real-time on-street parking detection and visualization on an edge device
KR102484789B1 (en) Intelligent crossroad integration management system with unmanned control and traffic information collection function
Pan et al. Identifying Vehicles Dynamically on Freeway CCTV Images through the YOLO Deep Learning Model.
Yeh et al. Detection and Recognition of Arrow Traffic Signals using a Two-stage Neural Network Structure.
CN112537301B (en) Driving reference object selection method and device for intelligent driving traffic carrier
CN114913447A (en) Police intelligent command room system and method based on scene recognition
CN112634610A (en) Natural driving data acquisition method and device, electronic equipment and storage medium
ElHakim et al. Traffisense: A smart integrated visual sensing system for traffic monitoring
CN112528787A (en) Signal lamp fault detection method based on deep learning
KR102340902B1 (en) Apparatus and method for monitoring school zone

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20210713

RJ01 Rejection of invention patent application after publication