Automatic identification method for traffic incident

By dividing the algorithm tasks in traffic scenarios into classification, detection and tracking, and combining video timing information, using Densnet121, YoloX and FairMOT algorithms for traffic event recognition, the problem of low recognition accuracy in the existing technology is solved, and traffic is realized The rapid and accurate identification of events provides a scientific basis for traffic management.

CN119091399AInactive Publication Date: 2024-12-06YINGQIANJIA INTELLIGENT TECH (WUHAN) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411090106.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-09
Publication Date
2024-12-06
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing image-based traffic event recognition technology fails to effectively consider video timing information, resulting in low recognition accuracy and difficulty in implementing it in practical applications.

Method used

The algorithmic tasks in traffic scenarios are divided into three categories: classification, detection and tracking, and the identification process and judgment logic of different traffic events are proposed based on the algorithm results. The video data is processed using Densnet121, YoloX and FairMOT algorithms, and the traffic events are processed in combination with video timing information. automated identification.

Benefits of technology

It realizes rapid identification of traffic events, provides scientific decision-making basis, effectively reduces the impact of accidents, and improves the accuracy and efficiency of traffic management.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119091399A_ABST
    Figure CN119091399A_ABST
Patent Text Reader

Abstract

The invention relates to the field of intelligent traffic, and discloses an automatic identification method for traffic events. According to an actual traffic scene, algorithm tasks in the traffic scene are divided into classification, detection and tracking, identification processes and judgment logics of different traffic events are provided according to algorithm results, and automatic identification of various traffic events is realized. The traffic accident abnormal condition can be quickly identified, a scientific decision-making basis can be provided for a traffic management department, and the influence of the accident can be effectively reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation, and in particular to an automatic identification method for traffic events. Background Art

[0002] With the continuous development of science and technology, the technology of automatic identification of traffic incidents is also constantly improving. At present, intelligent traffic identification technology has been widely used and has become an indispensable part of modern traffic management. The identification of various traffic incidents can not only serve as a means of auxiliary emergency response, but also is an important guarantee for improving traffic efficiency and traffic safety, and plays an important role in promoting the sustainable and healthy development of the transportation industry.

[0003] However, although these technologies have made some progress, they may still face many challenges in practical applications. For example, based on image information, the temporal information of the video is not considered, which makes the actual traffic event identification unfeasible or difficult to guarantee accuracy, which makes it difficult to implement in practical applications. Summary of the invention

[0004] In order to solve at least one of the problems mentioned in the above background technology, the present invention proposes an automatic identification method for traffic events.

[0005] A method for automatically identifying traffic incidents, specifically, the workflow includes the following steps:

[0006] Step S1, the image acquisition device acquires video data and sends the acquired video data to the computing node;

[0007] Step S2, the computing node receives the video sent by the image acquisition device and processes the video;

[0008] Step S3, performing calculations on the data processed by the computing node to obtain the recognition result of the traffic event. Specifically, the processed data includes classification algorithm inference data, detection algorithm inference data, and tracking algorithm inference data, wherein:

[0009] The classification algorithm inference result is the inference result Result_class output by Densnet121. The format of Result_class is:

[0010] Result_class={"Camer_ID": Cameranumber, "Cls": category, "Score": Confidence}

[0011] Among them, Camer_ID is the camera number, Cls is the classification category, and Score is the confidence of the classification;

[0012] The detection algorithm inference result is the inference result Result_detect output by YoloX. The format of Result_detect is:

[0013] Result_detect={"Camer_ID": Cameranumber, "Cls": category, "Score": Confidence,

[0014] "detect_rect": (x1, y1), (x2, y2)}

[0015] Among them, Camer_ID is the camera number, Cls is the classification category, Score is the classification confidence, detect_rect is the coordinate position of the detection box, x1 and y1 are the coordinate values ​​of the X axis and Y axis of the upper left corner of the detection box, respectively, and x2 and y2 are the coordinate values ​​of the X axis and Y axis of the lower right corner of the detection box, respectively;

[0016] The tracking algorithm inference result is the inference result Result_track output by FairMOT. The format of Result_track is:

[0017] Result_track={"Camer_ID": Camera number, "R_ID": Re identification number, "track_rec": (x1, y1)}

[0018] Among them, Camer_ID is the camera number, R_ID is the re-identification number, Score is the confidence of the classification, track_rec is the coordinate position of the track point, x1 and y1 represent the coordinate value of the X axis and Y axis of the track point respectively;

[0019] Although automatic traffic identification technology has made certain progress, it may still face many challenges in practical applications. For example, based on image information, the timing information of the video is not considered, which makes the actual traffic event identification lack of feasibility, or it is difficult to ensure accuracy, which makes it difficult to implement in practical applications. According to the actual traffic scene, the present invention divides the algorithm tasks in the traffic scene into three categories: classification, detection and tracking, and proposes the identification process and judgment logic of different traffic events based on the algorithm results, realizing the automatic identification of various traffic events.

[0020] The traffic events include: illegal lane change by motor vehicles, traffic congestion, low visibility, non-motor vehicles breaking the ban, and motor vehicles occupying the emergency lane.

[0021] The judgment condition for illegal lane change of a motor vehicle is that the trajectory line of the motor vehicle intersects with the lane line. The specific calculation formula is:

[0022] a i x+b i =a j x+b j

[0023] Among them, a i x+b i is the trajectory of the vehicle, a j x+b j is the trajectory of the lane. When the above equation is established and there is a solution, it is determined that the motor vehicle changes lanes.

[0024] Vehicle congestion requires the following conditions to be met:

[0025] Condition N1: The number of motor vehicles detected in the same image is greater than the set threshold. The specific calculation formula is:

[0026] Num≥T1

[0027] Condition N2, Num is the number of vehicles, T1 is the set threshold;

[0028] Condition N3: The speed of the motor vehicle is less than the set speed threshold. The specific calculation formula is:

[0029] S≥T s

[0030]

[0031] Where S is the speed of the vehicle, T s is the set speed threshold, t j and t i represents the different times corresponding to the inference of image j and image i, y j and i Represents the y-axis coordinate of the same vehicle trajectory point in image i and image j.

[0032] Low visibility is obtained according to the following steps:

[0033] First, the images of each highway section are divided into four levels according to the visibility: extremely good, extremely average, extremely poor, and extremely poor.

[0034] Then, the classified images are divided into training set and validation set, and the Densnet121 classifier is trained;

[0035] Next, the trained classifier is loaded into the edge computing node for inference to obtain the inference result;

[0036] Finally, if the visibility of the inference result is extremely poor or extremely poor, it is determined that the visibility of the current road section is low.

[0037] Non-motor vehicles must meet the following conditions to enter the ban:

[0038] First, non-motor vehicles and people are detected simultaneously;

[0039] Secondly, there is an intersection between the human target detection frame and the non-motor vehicle detection frame. The specific calculation formula is:

[0040]

[0041] iou≥0

[0042] Among them, iou represents the ratio of the intersection and union of the person target detection frame and the non-motor vehicle detection frame, Person and Vehicle represent the person target detection frame and the non-motor vehicle detection frame respectively, ∩ represents the intersection, ∪ represents the union, and if iou is greater than 0, it is determined that the person target detection frame and the non-motor vehicle detection frame are intersected;

[0043] The judgment condition for a motor vehicle to occupy the emergency lane is that the X-axis coordinate of the motor vehicle is within the emergency lane line. The specific calculation formula is:

[0044] X i -X j ≥T l

[0045] Among them, X i is the X-axis coordinate of the vehicle trajectory point, X j is the X-axis coordinate of the emergency lane line, T l is the set distance threshold, when X i -X j Less than the set distance threshold T l When the emergency lane is reached, it is determined that the motor vehicle is occupying the emergency lane.

[0046] The present invention proposes an automatic identification method for traffic incidents, which has the following beneficial effects compared with the existing technology:

[0047] According to the actual traffic scene, the present invention divides the algorithm tasks in the traffic scene into three categories: classification, detection and tracking, and proposes the identification process and judgment logic of different traffic events according to the algorithm results, realizing the automatic identification of various traffic events. It can quickly identify abnormal situations of traffic accidents, provide scientific decision-making basis for traffic management departments and effectively reduce the impact of accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose and features of the present invention more obvious and easy to understand, the present technical solution is described in detail below through embodiments and in conjunction with the accompanying drawings.

[0050] like Figure 1 As shown, a method for automatically identifying traffic incidents, the workflow includes the following steps:

[0051] Step S1: an image acquisition device acquires video data and sends the acquired video data to a computing node.

[0052] Step S2: the computing node receives the video sent by the image acquisition device and processes the video.

[0053] Step S3, performing calculations on the data processed by the computing node to obtain the recognition result of the traffic event. Specifically, the processed data includes classification algorithm inference data, detection algorithm inference data, and tracking algorithm inference data, wherein:

[0054] The classification algorithm inference result is the inference result Result_class output by Densnet121. The format of Result_class is:

[0055] Result_c1ass={"Camer_ID": Camera number, "Cls": category, "Score": Confidence}

[0056] Among them, Camer_ID is the camera number, Cls is the classification category, and Score is the confidence of the classification.

[0057] The detection algorithm inference result is the inference result Result_detect output by YoloX. The format of Result_detect is:

[0058] Result_detect={"Camer_ID": Camera number, "Cls": category, "Score": Confidence,

[0059] "detect_rect": (x1, y1), (x2, y2)}

[0060] Among them, Camer_ID is the camera number, Cls is the classification category, Score is the classification confidence, detect_rect is the coordinate position of the detection box, x1 and y1 are the coordinate values ​​of the X-axis and Y-axis of the upper left corner of the detection box, respectively, and x2 and y2 are the coordinate values ​​of the X-axis and Y-axis of the lower right corner of the detection box, respectively.

[0061] The tracking algorithm inference result is the inference result Result_track output by FairMOT. The format of Result_track is:

[0062] Result_track={"Camer_ID": Camera number, "R_ID": Re identification number, "track_rec": (x1, y1)}

[0063] Among them, Camer_ID is the camera number, R_ID is the re-identification number, Score is the confidence of the classification, track_rec is the coordinate position of the track point, x1 and y1 represent the coordinate value of the X-axis and Y-axis of the track point respectively.

[0064] The traffic events include: illegal lane change by motor vehicles, traffic congestion, low visibility, non-motor vehicles breaking the ban, and motor vehicles occupying the emergency lane.

[0065] The judgment condition for illegal lane change of a motor vehicle is that the trajectory line of the motor vehicle intersects with the lane line. The specific calculation formula is:

[0066] a i x+b i =a j x+b j

[0067] Among them, a i x+b i is the trajectory of the vehicle, a j x+b j is the trajectory of the lane. When the above equation is established and there is a solution, it is determined that the motor vehicle changes lanes.

[0068] Vehicle congestion requires the following conditions to be met:

[0069] Condition N1: The number of motor vehicles detected in the same image is greater than the set threshold. The specific calculation formula is:

[0070] Num≥T1

[0071] Condition N2, Num is the number of vehicles, and T1 is the set threshold.

[0072] Condition N3: The speed of the motor vehicle is less than the set speed threshold. The specific calculation formula is:

[0073] S≥T s

[0074]

[0075] Where S is the speed of the vehicle, T s is the set speed threshold, t j and t i represents the different times corresponding to the inference of image j and image i, y j and i Represents the y-axis coordinate of the same vehicle trajectory point in image i and image j.

[0076] Low visibility is obtained according to the following steps:

[0077] First, the acquired images of each highway section are divided into four levels according to the visibility: extremely good, extremely average, extremely poor, and extremely poor.

[0078] Then, the classified images are divided into training set and validation set, and the Densnet121 classifier is trained.

[0079] Next, the trained classifier is loaded into the edge computing node for inference to obtain the inference result.

[0080] Finally, if the visibility of the inference result is extremely poor or extremely poor, it is determined that the visibility of the current road section is low.

[0081] Non-motor vehicles must meet the following conditions to enter the ban:

[0082] First, non-motor vehicles and people are detected simultaneously;

[0083] Secondly, there is an intersection between the human target detection frame and the non-motor vehicle detection frame. The specific calculation formula is:

[0084]

[0085] iou≥0

[0086] Among them, iou represents the ratio of the intersection and union of the person target detection frame and the non-motor vehicle detection frame, Person and Vehicle represent the person target detection frame and the non-motor vehicle detection frame respectively, ∩ represents the intersection, ∪ represents the union, and if iou is greater than 0, it is determined that there is an intersection between the person target detection frame and the non-motor vehicle detection frame.

[0087] The judgment condition for a motor vehicle to occupy the emergency lane is that the X-axis coordinate of the motor vehicle is within the emergency lane line. The specific calculation formula is:

[0088] X i -X j ≥T l

[0089] Among them, X i is the X-axis coordinate of the vehicle trajectory point, Xj is the X-axis coordinate of the emergency lane line, T l is the set distance threshold, when X i -X j Less than the set distance threshold T l When the emergency lane is reached, it is determined that the motor vehicle is occupying the emergency lane.

[0090] So far, according to the method disclosed in the present invention, the working process of the present invention has been implemented once.

[0091] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0092] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0093] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for automatically identifying traffic incidents, characterized in that: Includes steps: Step S1, the image acquisition device acquires video data and sends the acquired video data to the computing node; Step S2, the computing node receives the video sent by the image acquisition device and processes the video; Step S3, performing calculations on the data processed by the computing nodes to obtain a recognition result of the traffic event.

2. According to the method for automatic identification of traffic incidents in claim 1, the step S3 of performing calculations on the data processed by the computing nodes includes classification algorithm inference data, detection algorithm inference data, and tracking algorithm inference data, wherein: The classification algorithm inference result is the inference result Result_class output by Densnet121. The format of Result_class is: Result_class = {"Camer_ID": Camera number, "Cls": category, "Score": Confidence}, where Camer_ID is the camera number, Cls is the category, and Score is the confidence of the classification; The detection algorithm inference result is the inference result Result_detect output by YoloX. The format of Result_detect is: Result_detect="{"Camer_ID": Camera number, "Cls": category, "Score": Confidence, "detect_rect": (x1, y1), (x2, y2)} Among them, Camer_ID is the camera number, Cls is the classification category, Score is the classification confidence, detect_rect is the coordinate position of the detection box, x1 and y1 are the coordinate values ​​of the X axis and Y axis of the upper left corner of the detection box, respectively, and x2 and y2 are the coordinate values ​​of the X axis and Y axis of the lower right corner of the detection box, respectively; The tracking algorithm inference result is the inference result Result_track output by FairMOT. The format of Result_track is: Result_track={"Camer_ID": Camera number, "R_ID": Re identification number, "track_rec": (x1, y1)} Among them, Camer_ID is the camera number, R_ID is the re-identification number, Score is the confidence of the classification, track_rec is the coordinate position of the track point, x1 and y1 represent the coordinate value of the X axis and Y axis of the track point respectively; The traffic events obtained include: illegal lane changes by motor vehicles, traffic congestion, low visibility, non-motor vehicles violating traffic rules, and motor vehicles occupying emergency lanes.

3. According to the method for automatically identifying traffic events in claim 2, the judgment condition for illegal lane change of the motor vehicle is that the trajectory line of the motor vehicle intersects with the lane line, and the specific calculation formula is: a i x+b i =a j x+b j in, a i x+b i is the trajectory of the vehicle, a j x+b j is the trajectory of the lane. When the above equation is established and there is a solution, it is determined that the motor vehicle changes lanes.

4. According to the method for automatically identifying traffic events in claim 2, the vehicle congestion needs to meet the following conditions at the same time: Condition N1: The number of motor vehicles detected in the same image is greater than the set threshold. The specific calculation formula is: Num≥T1 Condition N2, Num is the number of vehicles, T1 is the set threshold; Condition N3: The speed of the motor vehicle is less than the set speed threshold. The specific calculation formula is: S≥T s in, S is the speed of the vehicle, T s is the set speed threshold, t j and t i represents the different times corresponding to the inference of image j and image i, y j and i Represents the y-axis coordinate of the same vehicle trajectory point in image i and image j.

5. The method for automatically identifying traffic events according to claim 2, wherein the low visibility is obtained according to the following steps: Step A1, classifying each acquired highway section image into four levels according to the visibility: extremely good, extremely average, extremely poor, and extremely poor; Step A2, divide the classified images into training set and validation set, and train the Densnet121 classifier; Step A3: Load the trained classifier to the edge computing node for inference to obtain the inference result; Step A4: If the visibility of the inference result is extremely poor or extremely poor, it is determined that the visibility of the current road section is low.

6. According to the method for automatically identifying a traffic incident in claim 2, the non-motor vehicle breaking the ban needs to meet the following conditions: Condition M1, non-motor vehicle targets and human targets are detected simultaneously; Condition M2: There is an intersection between the human target detection frame and the non-motor vehicle detection frame. The specific calculation formula is: in, iou represents the ratio of the intersection and union of the person target detection frame and the non-motor vehicle detection frame. Person and Vehicle represent the person target detection frame and the non-motor vehicle detection frame respectively. ∩ represents the intersection and ∪ represents the union. If iou is greater than 0, it is determined that there is an intersection between the person target detection frame and the non-motor vehicle detection frame.

7. According to the method for automatically identifying traffic events in claim 2, the judgment condition for a motor vehicle to occupy an emergency lane is that the X-axis coordinate of the motor vehicle is located within the emergency lane line, and the specific calculation formula is: X i -X j ≥T l in, X i is the X-axis coordinate of the vehicle trajectory point, X j is the X-axis coordinate of the emergency lane line, T l is the set distance threshold, when X i -X j Less than the set distance threshold T l When the emergency lane is reached, it is determined that the motor vehicle is occupying the emergency lane.