Traffic safety detection method and system based on car-following distance

Abstract

The application discloses a traffic safety detection method and system based on a following distance, and relates to the field of intelligent traffic. The method comprises the following steps: collecting multi-angle sensing data of N vehicles based on a multi-angle data sensing device group arranged in a target area; taking image information of the N vehicles as a decomposition target, traversing multi-angle video sensing data to perform target decomposition, and generating multi-angle vehicle video sensing data; constructing a three-dimensional point cloud space of the target area; performing target decomposition on multi-angle radar sensing data to generate multi-angle vehicle radar sensing data; obtaining M following distance data; performing safety evaluation, and performing safety detection on traffic in the target area according to a safety evaluation result. The technical problems of insufficient accuracy and poor environmental adaptability of the existing following distance detection are solved, and the technical effects of improving the accuracy and environmental adaptability of the following distance detection are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent transportation, and particularly relates to a traffic safety detection method and system based on following distance. BACKGROUND

[0002] With the rapid development of transportation industry, road traffic safety problems are increasingly prominent. Following distance detection is crucial to ensure the safety and smoothness of road traffic. Maintaining appropriate following distance can reduce the risk of rear-end accidents, because when the distance between vehicles is too close, sudden deceleration or stopping of the front vehicle can cause the rear vehicle to have no time to react and result in rear-end collision. In addition, appropriate following distance can also provide enough reaction time for the driver to deal with sudden traffic conditions, such as sudden lane change of the front vehicle or pedestrian crossing the road suddenly, etc. Therefore, accurate detection of following distance is of great significance to improve the safety of road traffic and reduce the accident rate. Existing following distance detection methods mainly rely on single sensors such as radar or camera. These methods estimate the following distance by measuring the distance or relative speed between vehicles. However, due to the influence of various environmental factors such as bad weather, light conditions, etc., the detection of following distance by single sensor is not accurate, in addition, single sensor cannot cope with complex traffic scenarios such as multi-lane, intersection, etc., resulting in poor environmental adaptability

[0003] In the related art, the following distance detection has the technical problems of insufficient accuracy and poor environmental adaptability. SUMMARY

[0004] The present application provides a traffic safety detection method and system based on following distance, which adopts multiple angle data sensing device groups, constructs a three-dimensional point cloud space of the target area, and fuses multiple angle sensing data, etc. technical means, which achieves the technical effect of improving the accuracy and environmental adaptability of following distance detection.

[0005] The present application provides a traffic safety detection method based on following distance, comprising:

[0006] Based on the multiple angle data sensing device groups arranged in the target area, multiple angle sensing data of N vehicles in the target area are collected, wherein the multiple angle sensing data can include multiple angle video sensing data and multiple angle radar sensing data;

[0007] Taking the image information of the N vehicles as the decomposition target, the multiple angle video sensing data is traversed for target decomposition to generate multiple angle vehicle video sensing data;

[0008] Based on the multiple angle vehicle video sensing data, a three-dimensional point cloud space of the target area is constructed;

[0009] According to the three-dimensional point cloud space, the multi-angle radar sensing data is target-decomposed to generate multi-angle vehicle radar sensing data;

[0010] Through analyzing the multi-angle vehicle radar sensing data, M car-following distance data of N vehicles in the target area are obtained;

[0011] Based on the M car-following distance data, the safety of N vehicles is evaluated, and the traffic in the target area is detected according to the safety evaluation result.

[0012] The application also provides a traffic safety detection system based on car-following distance, comprising:

[0013] A multi-angle sensing data acquisition module is configured to acquire multi-angle sensing data of N vehicles in a target area based on a multi-angle data sensing device group arranged in the target area, wherein the multi-angle sensing data can include multi-angle video sensing data and multi-angle radar sensing data.

[0014] A multi-angle vehicle video sensing data generation module is configured to take image information of N vehicles as a decomposition target, to traverse the multi-angle video sensing data for target decomposition, and to generate multi-angle vehicle video sensing data.

[0015] A three-dimensional point cloud space construction module is configured to construct a three-dimensional point cloud space of the target area based on the multi-angle vehicle video sensing data.

[0016] A multi-angle vehicle radar sensing data generation module is configured to target-decompose the multi-angle radar sensing data according to the three-dimensional point cloud space to generate multi-angle vehicle radar sensing data.

[0017] A car-following distance data acquisition module is configured to obtain M car-following distance data of N vehicles in the target area by analyzing the multi-angle vehicle radar sensing data.

[0018] A traffic safety detection module is configured to evaluate the safety of N vehicles based on the M car-following distance data, and to detect the traffic in the target area according to the safety evaluation result.

[0019] The traffic safety detection method and system based on the following distance of the vehicle provided in the application can first collect multi-angle sensing data of N vehicles in a target area based on a multi-angle data sensing device group arranged in the target area, wherein the multi-angle sensing data can include multi-angle video sensing data and multi-angle radar sensing data, then image information of the N vehicles is taken as a decomposition target, multi-angle video sensing data is traversed for target decomposition, multi-angle vehicle video sensing data is generated, a three-dimensional point cloud space of the target area is constructed based on the multi-angle vehicle video sensing data, then multi-angle radar sensing data is decomposed according to the three-dimensional point cloud space, multi-angle vehicle radar sensing data is generated, M following distance data of the N vehicles in the target area are obtained by analyzing the multi-angle vehicle radar sensing data, finally, the N vehicles are evaluated based on the M following distances, and traffic in the target area is detected based on the evaluation result, thereby achieving the technical effects of improving the accuracy and environmental adaptability of the following distance detection. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings of the embodiments of the application will be briefly introduced below, and the flowchart is used to illustrate the operations performed by the system according to the embodiments of the application in the present application. It should be understood that the foregoing or the following operations are not necessarily performed in sequence. On the contrary, various steps can be processed in reverse order or simultaneously as needed. Meanwhile, other operations can be added to these processes, or one or more steps of operations can be removed from these processes.

[0021] Figure 1 The flowchart of the traffic safety detection method based on the following distance of the vehicle provided in the embodiments of the application is shown in the figure.

[0022] Figure 2 The structure diagram of the traffic safety detection system based on the following distance of the vehicle provided in the embodiments of the application is shown in the figure. DETAILED DESCRIPTION

[0023] The above description is only a summary of the technical solutions of the application, in order to more clearly understand the technical means of the application, the content of the specification can be implemented, and in order to make the above and other purposes, features and advantages of the application more obvious and easy to understand, the following specific embodiments of the application are described.

[0024] In order to make the purpose, technical scheme and advantages of the application more clear, the application will be further described in detail below in combination with the drawings, and the described embodiments should not be regarded as limiting the application, all other embodiments obtained by the person skilled in the art without making creative labor belong to the protection scope of the application.

[0025] In the following description, "some embodiments" are referred, which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subset of all possible embodiments, and can be combined with each other without conflict, the term "first\second" referred to only distinguishes similar objects, and does not represent a specific order for the objects. The terms "include" and "have" and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as understood by those skilled in the art to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application.

[0026] The embodiments of the present application provide a traffic safety detection method based on following distance, as shown in the following table: Figure 1 The method comprises the following steps:

[0027] In step S100, multi-angle sensing data of N vehicles in a target area is collected based on a multi-angle data sensing device group arranged in the target area, wherein the multi-angle sensing data can include multi-angle video sensing data and multi-angle radar sensing data. The target area refers to a specific area range to be detected for traffic safety, such as a section of road, a road intersection, a parking lot, etc. The multi-angle data sensing device group is deployed in the target area, including cameras, radar sensors, lidar, etc. These devices are placed in positions that can capture multi-angle information of vehicles, such as along the sides of the road, traffic light poles, buildings, etc. The multi-angle data sensing device group is used to continuously or periodically collect multi-angle sensing data of N (N is a positive integer, N≥2) vehicles in the target area, wherein the multi-angle video sensing data is obtained by a camera, the multi-angle radar sensing data is obtained by a radar device, and the multi-angle sensing data includes information such as the position, speed, direction and size of the vehicle.

[0028] Step S200, taking the image information of N vehicles as the decomposition target, traversing the multi-angle video sensing data to perform target decomposition, and generating multi-angle vehicle video sensing data. Specifically, first, the collected multi-angle video sensing data is preprocessed, including denoising, image enhancement and the like, and then a target detection algorithm in computer vision technology, such as YOLO (You Only Look Once), SSD (Single Shot MultiBox Detector) or Faster R-CNN (Region Convolutional Neural Networks) and the like, is used to identify the position and bounding box of the vehicle in the video frame. By aligning, splicing or fusing the image information of different angles, the video sensing data collected from different angles is fused to generate the multi-angle vehicle video sensing data, which reflects more comprehensive visual information of the vehicle. The generated multi-angle vehicle video sensing data is labeled and verified, and the labeling content includes the type, color, license plate number and the like of the vehicle.

[0029] In a possible implementation, step S200 further includes step S210, taking the image information of N vehicles as the decomposition target, performing video segmentation on the multi-angle video sensing data based on a shot boundary detection algorithm to obtain a plurality of image frame information. Using a shot boundary detection algorithm, such as a method based on histogram comparison, pixel statistics or other advanced algorithms, to identify the shot changes in the video, the shot boundary marks the change of video content, such as scene switching, vehicle appearance or disappearance, etc. After detecting the shot boundary, the video is segmented into a plurality of independent segments or frame sequences, each of which contains relatively consistent content, such as the continuous driving process of a single vehicle. The image frame information is extracted from each segmented video segment. Step S220, extracting key frames from the plurality of image frame information to obtain Q key frames. First, define the selection criteria for key frames, including the clarity of the frame, the integrity of the vehicle appearance, the significance of the feature change, etc. Using frame difference method, histogram comparison and other methods to compare the differences between adjacent frames frame by frame, according to the comparison result and the key frame selection criteria, Q key frames are extracted from the plurality of image frame information, which represent the main content of the video, reduce data redundancy, and at the same time retain sufficient information. Step S230, combining the Q key frames to construct multi-angle vehicle video sensing data. That is, using image registration, alignment and fusion technology to fuse the key frames obtained from different angles and sensors, and integrating the fused key frame data into a data set, i.e. the multi-angle vehicle video sensing data, which contains the position, posture, motion trajectory and other information of the vehicle under different angles. This implementation effectively extracts key information from multi-angle video sensing data, achieving the technical effect of constructing more concise and representative multi-angle vehicle video sensing data.

[0030] At step S300, a three-dimensional point cloud space of the target area is constructed based on the multi-angle vehicle video sensing data. Specifically, depth information is extracted from the multi-angle vehicle video sensing data using stereo vision technology. Two or more cameras are used to capture the same scene from different angles, and then the depth is calculated by matching feature points in different views. Then, a three-dimensional reconstruction algorithm such as SFM (Structure from Motion) or SLAM (Simultaneous Localization and Mapping) is applied to extract three-dimensional point cloud data from the video frames. The three-dimensional reconstruction algorithm reconstructs the three-dimensional structure of the scene by analyzing the motion of feature points in consecutive image frames. Next, the three-dimensional point cloud data obtained from different angles and sensors is registered and fused through point cloud data alignment and coordinate transformation to form a unified three-dimensional point cloud space. The constructed three-dimensional point cloud space is optimized and filtered to remove noise points, redundant data and smooth surfaces, improving the quality and usability of the point cloud data. Based on the optimized point cloud data, a three-dimensional model of the target area is constructed, which contains accurate geometric information of the vehicle.

[0031] In one possible implementation, step S300 further includes step S310 of performing feature extraction based on the Q key frames of the multi-angle vehicle video sensing data to obtain multi-angle vehicle feature values. That is, the vehicle features to be extracted are determined, which can be the contour, texture, corner, edge, etc. of the vehicle. Computer vision algorithms such as SIFT, SURF, ORB, etc. are applied to extract these features from each key frame, and the extracted features are converted into numerical representations, i.e. multi-angle vehicle feature values. At step S320, the multi-angle vehicle feature values are mapped and matched to obtain feature association information. Feature matching algorithms such as FLANN, BFMatcher, etc. are used to find similar feature points in multiple views, and the multi-angle vehicle feature values in different key frames are matched to establish association information between different key frames. At step S330, a three-dimensional reconstruction algorithm is used to construct the three-dimensional point cloud space based on the feature association information. The three-dimensional reconstruction algorithm can be a feature-based three-dimensional reconstruction or an optimization-based three-dimensional reconstruction, etc. The feature association information obtained in step S320 is used to calculate the three-dimensional position and direction of the vehicle under different views through the three-dimensional reconstruction algorithm, and the calculated three-dimensional position and direction information is converted into point cloud data to construct the three-dimensional point cloud space of the target area. This implementation accurately extracts vehicle features and establishes associations between key frames, achieving the technical effect of generating an accurate three-dimensional point cloud space, which provides more reliable basic data for subsequent radar data processing, car-following distance calculation and safety evaluation.

[0032] In one possible implementation, step S310 further includes step S311, extracting an I-frame image; and step S312, segmenting the I-frame image according to a preset segmentation standard to determine a target I-frame image block. First, the I-th frame (where I can be the index of any keyframe) is selected from Q keyframes as the object of feature extraction. Based on vehicle features and image characteristics, a preset segmentation standard is established, such as based on image size, vehicle outline, etc. According to the segmentation standard, the I-frame image is segmented to obtain multiple image blocks. Based on vehicle features (such as outline, color, etc.) and the segmentation results, a target I-frame image block containing vehicle features is determined. Step S313, discrete cosine transform coefficients are calculated based on the target I-frame image block to obtain a first eigenvalue and a second eigenvalue. The discrete cosine transform is applied to the target I-frame image block to convert the image from the spatial domain to the frequency domain. The formula for calculating the first eigenvalue is as follows:

[0033] T n =a∑ x′，y′ DC n (x′,y′)+b∑ x′，y′ AC n (x′, y′), (a+b=1, and a>b);

[0034] T n This refers to the first eigenvalue, DC. n (x′, y′) refers to the first DC coefficient, AC n (x′, y′) refers to the first AC coefficient, n refers to the first I-frame image, (x′, y′) refers to the (x′, y′)th sub-block of the first I-frame image, a refers to the influence factor of the first DC coefficient on the first eigenvalue, and b refers to the influence factor of the first AC coefficient on the first eigenvalue. The formula for calculating the second eigenvalue is as follows:

[0035] T n+1 =c∑ x″，y″ DC n+1 (x″,y″)+d∑ x″，y″ AC n+1 (x″,y″), (c+d=1,and,c>d);

[0036] T n+1 This refers to the second eigenvalue, DC. n+1 (x″, y″) refers to the second DC coefficient, AC n+1(x", y") is a second alternating coefficient, n+1 is a second I-frame image, (x", y") is an (x", y") sub-block of the second I-frame image, c is an influence factor of the second direct current coefficient on the second characteristic value, and d is an influence factor of the second alternating coefficient on the second characteristic value. In step S314, an I-frame image characteristic value is calculated according to the first characteristic value and the second characteristic value by using a characteristic value calculation formula. The characteristic value calculation formula can be a weighted sum, a difference, a ratio, etc. of two characteristic values, and the I-frame image characteristic value is calculated by using the defined specific characteristic value calculation formula. In step S315, the I-frame image characteristic value is added to the multi-angle vehicle characteristic value. That is, the calculated I-frame image characteristic value is added to the extracted multi-angle vehicle characteristic value set, and the characteristic values of the remaining key frames are calculated in the same way, so as to obtain a set containing the multi-angle vehicle characteristic values extracted from the Q key frames. This implementation mode accurately extracts the vehicle characteristics in the image and calculates the corresponding characteristic values, thereby achieving the technical effect of providing important basic information for subsequent feature matching and three-dimensional point cloud space construction.

[0037] In step S400, target decomposition is performed on the multi-angle radar sensing data according to the three-dimensional point cloud space, and multi-angle vehicle radar sensing data is generated. Specifically, the fusion of the three-dimensional point cloud space and the multi-angle radar sensing data is achieved by matching the vehicle positions in the multi-angle radar sensing data with the corresponding positions in the three-dimensional point cloud space. In the fused data, a target recognition algorithm (such as a clustering algorithm, a classifier, etc.) is used to recognize vehicle targets according to the shape, size, speed, etc. of the vehicles. For the recognized vehicle targets, target decomposition is performed, and the targets are segmented into different parts (such as front, rear, side, etc.) according to the shape, size, etc. of the vehicles. According to the target decomposition result, the multi-angle vehicle radar sensing data is generated, which includes the position, speed, shape, etc. of each vehicle part.

[0038] Step S500, by analyzing the multi-angle vehicle radar sensing data, obtaining M vehicle following distance data of N vehicles in the target area. First, pre-process the multi-angle vehicle radar sensing data, including removing noise, smoothing data, calibrating sensors, etc., to ensure the accuracy and consistency of the data. Next, use a vehicle tracking algorithm to identify and track vehicles in the target area, match the continuous radar scan data with the previous detection results, and identify the motion trajectory of each vehicle. For each tracked vehicle, calculate its following distance with the vehicle in front (if any). The following distance is defined as the shortest distance between the two vehicles in the same lane direction. Calculate the following distance by using the vehicle position information provided by the radar sensor. For example, if the radar sensor provides the precise position of the vehicle (such as latitude and longitude coordinates), the Euclidean distance between the two vehicles can be directly calculated. If the radar sensor provides the distance and angle information of the vehicle relative to the sensor, combine the position and direction information of the sensor, and use triangulation or other geometric methods to calculate the following distance. Since the vehicle radar sensing data is multi-angle, by fusing vehicle radar sensing data from different angles, the error can be reduced and the accuracy of the following distance can be improved. Finally, M (M is a positive integer, M≥1) vehicle following distance data of N vehicles in the target area is obtained, where M is less than N.

[0039] In a possible implementation, the step S500 further includes a step S510 of determining a detectable vehicle in the target area according to a laser beam emission limit of the laser radar. Parameters such as a detection range, an angle resolution, and a beam divergence angle of the laser radar are determined, and a range of vehicles that can be detected by the laser radar in the target area is calculated according to the detection range and the angle resolution of the laser radar. The range is a sector area with the laser radar as the center and the maximum detection distance as the radius. After the detectable area is determined, the vehicles located in the detectable area are identified as the detectable vehicles by analyzing the multi-angle vehicle radar sensing data. In step S520, a starting angle and an ending angle of the laser radar detection are obtained based on the detectable vehicles. In step S530, the laser beam is emitted and detected based on the starting angle and the ending angle, and the M car-following distance data of the N vehicles are determined. For each detectable vehicle, a bearing angle relative to the laser radar is calculated, which is the starting angle that needs to be adjusted when the laser radar emits the laser beam. According to the angle resolution and the detection requirement of the laser radar, a scanning range is set, which is an interval with the starting angle as the center and a certain angle as the width. The ending angle is the starting angle plus the interval width. According to the starting angle and the ending angle calculated in step S520, the emission angle of the laser radar is adjusted to be aligned with the target vehicle. After the angle is adjusted, the laser radar emits the laser beam to detect the target vehicle. After the laser beam is irradiated to the target vehicle, part of the beam is reflected back and received by the laser radar. By analyzing the time and intensity of the reflected signal, the time and distance experienced by the laser beam from emission to reception are calculated. According to the emission time, the reception time, and the speed of light of the laser beam, the distance between the target vehicle and the front vehicle, i.e., the car-following distance, is calculated based on the principles of triangulation or time difference ranging. Steps S520 and S530 are repeated for all the detectable vehicles in the target area to obtain the car-following distance data of each vehicle, and finally M car-following distance data of N vehicles are obtained. In this implementation, the laser radar has the characteristics of high precision and high reliability, can provide accurate distance and angle information, and can accurately calculate the car-following distance between the target vehicle and the front vehicle through the principles of triangulation or time difference ranging, thereby achieving the technical effect of obtaining high-precision car-following distance data.

[0040] At step S600, the N vehicles are evaluated for safety based on the M following distances, and the traffic in the target area is detected for safety according to the safety evaluation result. Specifically, the M following distance data are integrated to form a data set containing the following distance information of all vehicles in the target area. According to traffic regulations, safety standards or actual needs, one or more safety thresholds are set, which can be fixed or dynamic, adjusted according to traffic conditions, weather conditions and other factors. The safety threshold is set to determine which following distance is safe and which is dangerous. The following distance data of each vehicle is compared with the set safety threshold to determine whether the following distance of each vehicle is within the safe range. If the following distance is less than the safety threshold, it is considered that the vehicle has potential safety risks. Based on the safety evaluation result, the overall safety of the traffic in the target area is detected. If it is detected that there are multiple vehicles with close following distance, overspeed, frequent lane changing and other unsafe behaviors, it is considered that the traffic in the target area has potential safety risks. At this time, corresponding measures are taken for processing, such as reminding the driver, adjusting the timing of the traffic signal, guiding the vehicle to flow, etc., to improve the safety of the traffic. The embodiments of the present application adopt technical means such as laying out a multi-angle data sensing device group, constructing a three-dimensional point cloud space of the target area, and fusing multi-angle sensing data, to achieve the technical effects of improving the accuracy and environmental adaptability of the following distance detection.

[0041] In a possible implementation, the step S600 further includes a step S610 of setting a following safety distance interval based on vehicle driving rules. First, relevant traffic laws, vehicle manufacturer recommendations, road design standards, etc. are collected, which rules specify safe following distances under different roads and speeds. Based on the collected vehicle driving rules, a following safety distance interval is calculated using a mathematical model or algorithm. The following safety distance interval is calculated based on current vehicle speed, road conditions (such as wet, congestion, etc.), and driver reaction time, etc. According to the calculation result, the minimum and maximum values of the following safety distance are set as thresholds. In step S620, whether the M following distances are in the following safety distance interval is judged in turn. In the M following distance data, the following distance of each vehicle is compared with the set following safety distance interval to determine whether it is in this interval. For each vehicle, the result of whether the following distance meets the safety requirement is recorded. In step S630, if yes, the first safety marking of the tail and the head of the first group of adjacent vehicles with the following safety distance interval is performed. That is, the vehicles with the following distance in the following safety distance interval are identified from all vehicles, i.e. the first group of adjacent vehicles, and the first safety marking is performed at the tail and head positions of these adjacent vehicles. In step S640, if no, the first warning information is generated, wherein the first warning information is marked at the tail and the head of the first group of adjacent vehicles respectively. That is, the vehicles with the following distance not in the following safety distance interval are identified from all vehicles, the first warning information is generated for these vehicles, and is marked at the tail and head positions of these vehicles respectively. This implementation provides clear road traffic auxiliary information by marking safety or warning information at the head and tail of each vehicle, and achieves the technical effect of improving road traffic safety.

[0042] In a possible implementation, the step S600 further includes a step S650 of judging whether the head or tail of the first group of adjacent vehicles is within the following distance safety interval based on the first safety mark. That is, first, the first group of adjacent vehicles with the first safety mark is identified from all vehicles, for each vehicle in the first group, if the first safety mark is located at the tail, the following distance between the head and the front adjacent vehicle is obtained; if the first safety mark is located at the head, the following distance between the tail and the rear adjacent vehicle is obtained. According to the set following distance safety interval, the following distance is compared with the threshold value, and it is judged whether the obtained following distance data is within the following distance safety interval. If yes, the second safety mark is marked, and if no, the second warning information is generated. The N vehicles are iterated in this way, the safety evaluation is performed according to the iteration result, and the safety evaluation result is generated. Specifically, for the first group of adjacent vehicles whose following distance is within the following distance safety interval in the step S650, the second safety mark is marked at the head or tail position, and for the first group of adjacent vehicles whose following distance is not within the following distance safety interval in the step S650, the second warning information is generated and marked at the head or tail position. The steps S650 and S660 are applied to all N vehicles, that is, the head and tail of each vehicle are judged whether to mark the second safety mark or generate the second warning information according to the following distance with the adjacent vehicle. According to the iteration processing result of all vehicles, the final safety evaluation result is generated, which is a comprehensive index considering the following distance, speed, acceleration and other factors of all vehicles, and whether there is the first safety mark, the second safety mark or the warning information, which is used for traffic management, vehicle scheduling, driver training and other aspects. This implementation makes more detailed judgment on the following distance between vehicles, and provides more accurate safety mark and warning information, and achieves the technical effect of further improving the road traffic safety.

[0043] In the foregoing, the traffic safety detection method based on the following distance according to the embodiments of the present application is described in detail. Next, the traffic safety detection system based on the following distance according to the embodiments of the present application will be described with reference to the drawings. Figure 1 The traffic safety detection system based on the following distance according to the embodiments of the present application is described in detail. Next, the traffic safety detection system based on the following distance according to the embodiments of the present application will be described with reference to the drawings. Figure 2 The traffic safety detection system based on the following distance according to the embodiments of the present application is described in detail. Next, the traffic safety detection system based on the following distance according to the embodiments of the present application will be described with reference to the drawings.

[0044] The traffic safety detection system based on the following distance according to the embodiment of the present application is used to solve the technical problems of the existing following distance detection, such as the insufficient accuracy and poor environmental adaptability, and achieves the technical effect of improving the accuracy and environmental adaptability of the following distance detection. The traffic safety detection system based on the following distance comprises a multi-angle sensor data acquisition module 10, a multi-angle vehicle video sensor data generation module 20, a three-dimensional point cloud space construction module 30, a multi-angle vehicle radar sensor data generation module 40, a following distance data acquisition module 50, and a traffic safety detection module 60.

[0045] The multi-angle sensor data acquisition module 10 is used to acquire multi-angle sensor data of N vehicles in a target area based on a multi-angle data sensing device group arranged in the target area, wherein the multi-angle sensor data can include multi-angle video sensor data and multi-angle radar sensor data.

[0046] The multi-angle vehicle video sensor data generation module 20 is used to take image information of the N vehicles as a decomposition target, traverse the multi-angle video sensor data for target decomposition, and generate multi-angle vehicle video sensor data.

[0047] The three-dimensional point cloud space construction module 30 is used to construct a three-dimensional point cloud space of the target area based on the multi-angle vehicle video sensor data.

[0048] The multi-angle vehicle radar sensor data generation module 40 is used to perform target decomposition on the multi-angle radar sensor data according to the three-dimensional point cloud space, and generate multi-angle vehicle radar sensor data.

[0049] The following distance data acquisition module 50 is used to acquire M following distance data of the N vehicles in the target area by analyzing the multi-angle vehicle radar sensor data.

[0050] The traffic safety detection module 60 is used to perform safety evaluation on the N vehicles based on the M following distances, and perform safety detection on the traffic of the target area according to the safety evaluation result.

[0051] Below, the specific configuration of the multi-angle vehicle video sensing data generation module 20 will be described in detail. As described above, the image information of N vehicles is taken as the decomposition target, the multi-angle video sensing data is traversed for target decomposition, and the multi-angle vehicle video sensing data is generated. The multi-angle vehicle video sensing data generation module 20 can further include: a video segmentation unit for taking the image information of N vehicles as the decomposition target, performing video segmentation on the multi-angle video sensing data based on a shot boundary detection algorithm, and obtaining a plurality of image frame information; a key frame extraction unit for extracting key frames in the plurality of image frame information, and obtaining Q key frames; and a multi-angle vehicle video sensing data construction unit for constructing multi-angle vehicle video sensing data in combination with the Q key frames.

[0052] Below, the specific configuration of the three-dimensional point cloud space construction module 30 will be described in detail. As described above, the three-dimensional point cloud space of the target region is constructed based on the multi-angle vehicle video sensing data. The three-dimensional point cloud space construction module 30 can further include: a feature extraction unit for performing feature extraction based on the Q key frames of the multi-angle vehicle video sensing data, and obtaining multi-angle vehicle feature values; a mapping matching unit for mapping and matching the multi-angle vehicle feature values, and obtaining feature correlation information; and a three-dimensional point cloud space construction unit for constructing the three-dimensional point cloud space according to the feature correlation information by using a three-dimensional reconstruction algorithm.

[0053] In the feature extraction based on the Q key frames of the multi-angle vehicle video sensing data to obtain multi-angle vehicle feature values, the feature extraction unit can further include: an image segmentation subunit for extracting an I-frame image, segmenting the I-frame image according to a preset block standard, and determining a target I-frame image block; and a feature value calculation subunit for calculating discrete cosine transform coefficients based on the target I-frame image block, obtaining a first feature value and a second feature value, calculating an I-frame image feature value by using a feature value calculation formula according to the first feature value and the second feature value, and adding the I-frame image feature value to the multi-angle vehicle feature values.

[0054] Below, the specific configuration of the following vehicle distance data acquisition module 50 will be described in detail. As described above, the M following vehicle distance data of N vehicles in the target region are obtained by analyzing the multi-angle vehicle radar sensing data. The following vehicle distance data acquisition module 50 can further include: a detectable vehicle determination unit for determining detectable vehicles in the target region according to a laser beam emission limit of laser radar detection; a radar detection angle acquisition unit for obtaining a starting angle and a terminal angle of the laser radar detection based on the detectable vehicles; and a following distance determination unit for determining the M following vehicle distance data of N vehicles by performing emission detection on the laser beam based on the starting angle and the terminal angle.

[0055] The specific configuration of the traffic safety detection module 60 will be described in detail below. As described above, the traffic safety detection module 60 can further include a following safety distance interval setting unit configured to set a following safety distance interval based on a vehicle driving rule; a following distance judgment unit configured to judge whether the M following distances are in the following safety distance interval in turn; and a judgment result processing unit configured to, if yes, perform first safety marking of a tail and a head of a first group of adjacent vehicles having the following safety distance interval, and if no, generate first warning information, wherein the first warning information is marked on the tail and the head of the first group of adjacent vehicles respectively.

[0056] The judgment result processing unit can further include a safety evaluation sub-unit configured to, based on the first safety marking, preferentially judge whether a head or a tail of the first group of adjacent vehicles and a following distance of the remaining adjacent vehicles are in the following safety distance interval, if yes, perform second safety marking, if no, generate second warning information, and perform iteration on N vehicles, and perform safety evaluation based on the iteration result to generate the safety evaluation result.

[0057] The traffic safety detection system based on the following distance provided by the embodiments of the present application can perform the traffic safety detection method based on the following distance provided by any of the embodiments of the present application, and has the corresponding function modules and beneficial effects of performing the method.

[0058] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or the server, and the various units and modules are only divided according to the functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for the convenience of mutual differentiation, and do not limit the protection scope of the present application.

[0059] The above specific embodiments do not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application. In some cases, the actions or steps described in the present application can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

Claims

1. A traffic safety detection method based on car following distance, characterized in that, The method comprises: Based on the multi-angle data sensing device group arranged in the target area, collect multi-angle sensing data of N vehicles in the target area, wherein the multi-angle sensing data includes multi-angle video sensing data and multi-angle radar sensing data; Take the image information of the N vehicles as the decomposition target, traverse the multi-angle video sensing data for target decomposition, and generate multi-angle vehicle video sensing data; Construct a three-dimensional point cloud space of the target area based on the multi-angle vehicle video sensing data; According to the three-dimensional point cloud space, perform target decomposition on the multi-angle radar sensing data to generate multi-angle vehicle radar sensing data; By analyzing the multi-angle vehicle radar sensing data, obtain M vehicle following distance data of N vehicles in the target area; Based on the M vehicle following distance data, perform safety evaluation on the N vehicles, and perform safety detection on the traffic in the target area according to the safety evaluation result.

2. The car following distance based traffic safety detection method of claim 1, wherein, Take the image information of the N vehicles as the decomposition target, traverse the multi-angle video sensing data for target decomposition, and generate multi-angle vehicle video sensing data, which comprises: Take the image information of the N vehicles as the decomposition target, and perform video segmentation on the multi-angle video sensing data based on a lens boundary detection algorithm to obtain multiple image frame information; Extract key frames from the multiple image frame information to obtain Q key frames; Combine the Q key frames to construct multi-angle vehicle video sensing data. 3.The car following distance based traffic safety detection method of claim 2, wherein, Based on the Q key frames of the multi-angle vehicle video sensing data, perform feature extraction to obtain multi-angle vehicle feature values; Map and match the multi-angle vehicle feature values to obtain feature correlation information; Use a three-dimensional reconstruction algorithm to construct the three-dimensional point cloud space according to the feature correlation information. Based on the Q key frames of the multi-angle vehicle video sensing data, perform feature extraction to obtain multi-angle vehicle feature values, which comprises:

4. The car following distance based traffic safety detection method of claim 3, wherein, Extract I-frame images; Segment the I-frame images according to a preset block standard to determine target I-frame image blocks; Based on the target I-frame image blocks, calculate discrete cosine transform coefficients to obtain first feature values and second feature values; According to the first feature values and the second feature values, use a feature value calculation formula to calculate I-frame image feature values; Add the I-frame image feature values to the multi-angle vehicle feature values. By analyzing the multi-angle vehicle radar sensing data, obtain M vehicle following distance data of N vehicles in the target area, which comprises: 5.The car following distance based traffic safety detection method of claim 1, wherein, Determine the detectable vehicles in the target area according to the laser beam emission limit of the laser radar detection; Based on the detectable vehicles, obtain the start angle and end angle of the laser radar detection; Based on the start angle and the end angle, perform emission detection on the laser beam to determine the M vehicle following distance data of the N vehicles. It comprises: 6.The car following distance based traffic safety detection method according to claim 1, wherein the method Set a vehicle following safety distance interval based on vehicle driving rules; Judge whether the M vehicle following distances are in the vehicle following safety distance interval in sequence; ​ If yes, a first safety mark of the tail and the head of the first group of adjacent vehicles with the following safety distance interval is performed; If no, a first warning information is generated, wherein the first warning information is marked on the tail and the head of the first group of adjacent vehicles respectively.

7. The car following distance based traffic safety detection method as claimed in claim 6, wherein the method The method comprises: Based on the first safety mark, it is determined whether the following distance of the head or the tail of the first group of adjacent vehicles and the following distance of the remaining adjacent vehicles are in the following safety distance interval; If yes, a second safety mark is performed, if no, a second warning information is generated, and the iteration of N vehicles is performed, and the safety evaluation is performed according to the iteration result, and the safety evaluation result is generated.

8. A traffic safety detection system based on car following distance, characterized in that, The system is used to implement the traffic safety detection method based on the following distance according to any one of claims 1-7, and the system comprises: A multi-angle sensing data acquisition module, which is used to acquire multi-angle sensing data of N vehicles in a target area based on a multi-angle data sensing device group arranged in the target area, wherein the multi-angle sensing data comprises multi-angle video sensing data and multi-angle radar sensing data; A multi-angle vehicle video sensing data generation module, which is used to take image information of the N vehicles as a decomposition target, traverse the multi-angle video sensing data for target decomposition, and generate multi-angle vehicle video sensing data; A three-dimensional point cloud space construction module, which is used to construct a three-dimensional point cloud space of the target area based on the multi-angle vehicle video sensing data; A multi-angle vehicle radar sensing data generation module, which is used to perform target decomposition on the multi-angle radar sensing data according to the three-dimensional point cloud space, and generate multi-angle vehicle radar sensing data; A following distance data acquisition module, which is used to acquire M following distance data of the N vehicles in the target area by analyzing the multi-angle vehicle radar sensing data; A traffic safety detection module, which is used to perform safety evaluation on the N vehicles based on the M following distances, and perform safety detection on the traffic of the target area according to the safety evaluation result.

Images