Traffic accident scene analysis method and system based on multimodal data

Abstract

The present invention discloses a traffic accident scene analysis method and system based on multimodal data, which relates to the field of data analysis technology. The method comprises obtaining traffic accident scene information, obtaining image parameter information based on traffic accident scene image information, identifying targets based on the YOLO algorithm according to the traffic accident scene information, obtaining accident target information, and obtaining first accident target information and second accident target information according to the accident target information. The present invention determines the image recognition status through image recognition coefficients, providing a basis for subsequent image recognition settings. Based on image region classification information, image recognition of different specifications is performed on images of different regions, thereby ensuring image recognition accuracy while improving image recognition efficiency. The location of the traffic accident is reversed through accident target information to analyze the intensity of the traffic accident, thereby improving the efficiency and accuracy of traffic accident scene analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a traffic accident scene analysis method and system based on multimodal data. Background Art

[0002] At present, the analysis of traffic accident scenes still has problems such as the inability to accurately analyze traffic accident scene images, the inability to quickly identify accident targets, the inability to accurately analyze the accident situation based on the accident targets, and the inability to accurately evaluate traffic accidents. In existing technologies, accident targets are often identified through the YOLO algorithm, but the recognition accuracy is closely related to the number of grids. If the number of grids is too small, only the main body of the traffic accident, such as vehicles, can be identified, and small objects such as accident debris cannot be accurately identified. If the number of grids is too small, although the accident target can be accurately identified, the increase in the number of grids leads to a large amount of data processing, which reduces the efficiency of accident analysis. Summary of the Invention

[0003] In order to solve the above technical problems, a traffic accident scene analysis method and system based on multimodal data are provided. This technical solution solves the problems raised in the above background technology, such as the inability to accurately analyze traffic accident scene images, the inability to quickly identify accident targets, the inability to accurately analyze the accident situation based on the accident targets, and the inability to accurately evaluate traffic accidents. In the existing technology, accident targets are often identified through the YOLO algorithm, but the recognition accuracy is closely related to the number of grids. If the number of grids is too small, only the main body of the traffic accident, such as vehicles, can be identified, and small objects such as accident debris cannot be accurately identified. If the number of grids is too small, although the accident target can be accurately identified, the increase in the number of grids leads to a large amount of data processing, which reduces the efficiency of accident analysis.

[0004] In order to achieve the above objects, the technical solution adopted by the present invention is:

[0005] A traffic accident scene analysis method based on multimodal data, comprising:

[0006] Acquiring traffic accident scene information, wherein the traffic accident scene information includes traffic accident scene image information and scene map data;

[0007] Based on the traffic accident scene image information, image parameter information is obtained, wherein the image parameter information includes image resolution information and image size information;

[0008] According to the traffic accident scene information, target recognition is performed based on the YOLO algorithm to obtain accident target information;

[0009] Obtaining first accident target information and second accident target information according to the accident target information;

[0010] Acquiring characteristic trace information corresponding to a first accident target based on traffic accident scene image information, wherein the characteristic trace information represents brake mark information or skid mark information;

[0011] According to the first accident target, based on the characteristic traces, the accident target baseline is obtained;

[0012] According to the accident target baseline and the second accident target information, the traffic accident intensity is analyzed to obtain traffic accident information.

[0013] Preferably, the method of identifying the target based on the YOLO algorithm according to the traffic accident scene information to obtain the accident target information specifically includes:

[0014] According to the image parameter information, obtain image resolution information and image size information;

[0015] Based on the image resolution information, basic recognition parameter information is obtained, where the basic recognition parameter information represents the minimum length and minimum width of a recognizable object;

[0016] According to the basic recognition parameter information, the area of ​​the basic recognition rectangular region is obtained;

[0017] Obtain the image area of ​​the traffic accident scene according to the image size information;

[0018] The ratio of the area of ​​the basic recognition rectangle to the area of ​​the traffic accident scene image is used as the image recognition coefficient;

[0019] Obtaining a basic accident scene reference object and a reference object calibration group according to the image recognition coefficient, wherein the reference object calibration group includes a first reference object at the scene and a second reference object at the scene;

[0020] According to the reference object calibration group, the image acquisition height is obtained;

[0021] Obtain image area classification information based on basic accident scene reference objects and image acquisition height;

[0022] According to the image area classification information, the accident target is identified based on the YOLO algorithm to obtain accident target information. The accident target information includes a first accident target and a second accident target. The first accident target represents the main target of the traffic accident, and the second accident target represents the scattered objects after the traffic accident.

[0023] Preferably, obtaining the basic accident scene reference object and the reference object calibration group according to the image recognition coefficient specifically includes:

[0024] Acquiring accident scene reference object information based on traffic accident scene image information, wherein the accident scene reference object information includes image size information corresponding to each accident scene reference object, and the accident scene reference objects include lane lines, guardrails, and traffic signs;

[0025] Obtain the physical size information corresponding to the reference object at the accident scene, and use the ratio of the image size to the physical size as the image scale conversion coefficient;

[0026] The accident scene reference object corresponding to the maximum value of the image scale conversion coefficient is used as the first reference object at the scene;

[0027] The product of the maximum value of the image scale conversion coefficient and the image recognition coefficient is used as the image recognition coefficient threshold;

[0028] Screening accident scene reference objects according to an image recognition coefficient threshold to obtain basic accident scene reference objects;

[0029] If the image scale conversion coefficient corresponding to the accident scene reference object does not exceed the image recognition coefficient threshold, the accident scene reference object is removed;

[0030] The basic accident scene reference object corresponding to the minimum value of the image scale conversion coefficient among the basic accident scene reference objects is used as the second reference object at the scene;

[0031] A reference object calibration group is obtained according to the first on-site reference object and the second on-site reference object.

[0032] Preferably, obtaining the image acquisition height according to the reference object calibration group specifically includes:

[0033] According to the on-site map data, the first on-site reference object and the second on-site reference object are matched with the on-site map to obtain the position information of the calibration reference objects;

[0034] Acquire a reference object characteristic distance according to the calibration reference object position information, wherein the reference object characteristic distance represents a straight-line distance between a first reference object on site and a second reference object on site;

[0035] Obtaining a reference object conversion distance based on the traffic accident scene image information, wherein the reference object conversion distance represents a straight-line distance between two accident scene reference objects in a reference object calibration group in the traffic accident scene image;

[0036] Obtaining pixel size information and focal length information based on the traffic accident scene image information, wherein the pixel size represents the physical width corresponding to a single pixel;

[0037] The product of the reference object characteristic distance and the focal length is used as the spatial characteristic coefficient, and the product of the reference object conversion distance and the pixel size is used as the plane characteristic coefficient;

[0038] The ratio of the spatial characteristic coefficient to the plane characteristic coefficient is used as the image acquisition height.

[0039] Preferably, obtaining image area classification information based on basic accident scene reference objects and image acquisition height specifically includes:

[0040] The ratio of focal length to pixel size information is used as the image feature coefficient;

[0041] Obtaining a corresponding image scale conversion coefficient according to a first reference object and a second reference object on site;

[0042] The ratio of the image characteristic coefficient to the image scale conversion coefficient corresponding to the first reference object on the scene is used as the first straight-line distance, and the ratio of the image characteristic coefficient to the image scale conversion coefficient corresponding to the second reference object on the scene is used as the second straight-line distance;

[0043] According to the first straight-line distance and the image acquisition height, the first plane distance is obtained based on the Pythagorean theorem with the first straight-line distance as the hypotenuse of the triangle;

[0044] According to the second straight-line distance and the image acquisition height, the second plane distance is obtained based on the Pythagorean theorem with the second straight-line distance as the hypotenuse of the triangle;

[0045] Taking the first reference object on site as the center and the first plane distance as the radius, a first range curve is obtained;

[0046] Taking the second reference object on site as the center and the second plane distance as the radius, a second range curve is obtained;

[0047] The intersection of the first range curve and the second range curve is used as an image acquisition feature point;

[0048] Based on the basic accident scene reference objects, any one of the basic accident scene reference objects is used as the third reference object on site;

[0049] obtaining a third range curve according to a third reference object on site and an image characteristic coefficient;

[0050] The image acquisition feature point with the smallest distance from the third range curve is used as the image mapping point;

[0051] The distance between the image mapping point and the basic accident scene reference object is used as the image conversion reference value;

[0052] According to the image conversion reference value and the image scale conversion coefficient, the data is fitted with the image conversion reference value as the independent variable and the image scale conversion coefficient as the dependent variable to obtain the image scale characteristic coefficient;

[0053] Based on the image scale characteristic coefficient, an image scale conversion coefficient function is constructed;

[0054] Based on traffic accident target analysis, the minimum size and maximum size of the accident target standard are obtained;

[0055] Obtaining a first distance and a second distance according to a minimum size of an accident target standard, a maximum size of an accident target standard, and an image scale conversion coefficient function;

[0056] Divide the image by taking the image mapping point as the center and the first distance and the second distance as the region range to obtain image region classification information, wherein the image region classification information includes a first image region, a second image region, and a third image region;

[0057] The first distance and the second distance are specifically:

[0058]

[0059] Where, is the first distance, is the second distance, represents the image scale conversion coefficient corresponding to the first distance from the image mapping point, represents the image scale conversion coefficient corresponding to the second distance from the image mapping point, Indicates the standard minimum size of the accident target, Indicates the maximum size of the accident target standard, is the image scale conversion coefficient function, is the image scale characteristic coefficient, Indicates the image conversion reference value, Indicates the image acquisition height.

[0060] Preferably, analyzing the traffic accident intensity based on the accident target baseline and the second accident target information to obtain traffic accident information specifically includes:

[0061] Based on the analysis of traffic accident traces, the maximum deviation angle can be obtained by reverse estimating the traces;

[0062] According to the accident target baseline corresponding to the first accident target, the intersection of the accident target baseline is used as the basic accident occurrence point;

[0063] Taking the first accident target as the starting point, the accident target baseline is offset along both sides with half of the maximum deviation angle of the trace reverse estimation as the adjustment range, and the area formed by the basic accident occurrence point is regarded as the accident occurrence area;

[0064] obtaining second accident target density information according to the second accident target information;

[0065] The distance between each second accident target and the image mapping point is used as the image offset coefficient of the second accident target;

[0066] Obtaining an image scale conversion coefficient corresponding to each second accident target according to the image offset coefficient and the image scale conversion coefficient function;

[0067] The square of the image scale conversion coefficient is used as the image density correction coefficient corresponding to each second accident target;

[0068] The product of the second accident target density and the image density correction coefficient is used as the second accident target correction density;

[0069] The second accident target corresponding to the maximum value of the second accident target correction density is used as the accident base point;

[0070] The point closest to the accident base point in the accident area is taken as the accident occurrence point;

[0071] The sum of the distances between the first accident target and the accident occurrence point is used as the first accident intensity index;

[0072] acquiring a second accident target area according to the second accident target information;

[0073] The ratio of the area corresponding to each second accident target correction density to the second accident target area is used as the weight of the second accident target correction density;

[0074] Obtaining a second accident intensity index based on a weighted summation of the second accident target correction density and the corresponding weight;

[0075] The sum of the first accident intensity index and the second accident intensity index is used as the traffic accident intensity index;

[0076] Traffic accident information is obtained based on the first accident target information, the second accident target information, the accident occurrence point and the traffic accident intensity index.

[0077] Furthermore, a traffic accident scene analysis system based on multimodal data is proposed to implement the above analysis method, including:

[0078] A main control module, wherein the main control module is used to fit the data according to the image conversion reference value and the image scale conversion coefficient, with the image conversion reference value as the independent variable and the image scale conversion coefficient as the dependent variable, to obtain the image scale feature coefficient, to construct an image scale conversion coefficient function based on the image scale feature coefficient, to obtain the standard minimum size of the accident target and the standard maximum size of the accident target based on the traffic accident target analysis, to obtain the first distance and the second distance based on the standard minimum size of the accident target, the standard maximum size of the accident target and the image scale conversion coefficient function, to divide the image with the image mapping point as the center and the first distance and the second distance as the region range, to obtain image region classification information, to identify the accident target based on the image region classification information based on the YOLO algorithm, to obtain accident target information, and to obtain traffic accident information based on the accident target information;

[0079] an information acquisition module, the information acquisition module being used to acquire traffic accident scene information, the traffic accident scene information including traffic accident scene image information and scene map data, acquiring image parameter information based on the traffic accident scene image information, the image parameter information including image resolution information and image size information, acquiring basic recognition parameter information based on the image resolution information, the basic recognition parameter information indicating the minimum length and minimum width of a recognizable target, acquiring the area of ​​a basic recognition rectangular region based on the basic recognition parameter information, acquiring the image area of ​​the traffic accident scene based on the image size information, and acquiring reference object information at the accident scene based on the traffic accident scene image information;

[0080] An evaluation module, the evaluation module being configured to use the ratio of the area of ​​the basic identification rectangular area to the area of ​​the traffic accident scene image as an image recognition coefficient, obtain physical size information corresponding to the accident scene reference object, use the ratio of the image size to the physical size as an image scale conversion coefficient, obtain the basic accident scene reference object and the reference object calibration group based on the image recognition coefficient, obtain the corresponding first range curve and second range curve based on the first reference object and the second reference object at the scene, and obtain image mapping points based on the first range curve and the second range curve;

[0081] The display module interacts with the main control module and is used to output and display image area classification information, accident target information, accident target baseline, accident occurrence point and traffic accident information.

[0082] Optionally, the control unit is used to obtain the standard minimum size and the standard maximum size of the accident target based on traffic accident target analysis, obtain the first distance and the second distance according to the standard minimum size of the accident target, the standard maximum size of the accident target and the image scale conversion coefficient function, divide the image into regions with the image mapping point as the center and the first distance and the second distance as the region range, obtain image region classification information, identify the accident target based on the YOLO algorithm according to the image region classification information, obtain accident target information, and obtain traffic accident information according to the accident target information;

[0083] An information receiving unit, which interacts with the information acquisition module and the evaluation module to receive data and transmit it to the data processing unit;

[0084] The data processing unit is used to fit the data according to the image conversion reference value and the image scale conversion coefficient, with the image conversion reference value as the independent variable and the image scale conversion coefficient as the dependent variable, to obtain the image scale characteristic coefficient, and to construct an image scale conversion coefficient function based on the image scale characteristic coefficient.

[0085] Optionally, the information acquisition module specifically includes:

[0086] a first acquisition unit, configured to acquire traffic accident scene information, the traffic accident scene information including traffic accident scene image information and scene map data, and acquire image parameter information based on the traffic accident scene image information, the image parameter information including image resolution information and image size information;

[0087] The second acquisition unit is used to obtain basic recognition parameter information based on image resolution information, where the basic recognition parameter information represents the minimum length and minimum width of an identifiable target, obtain the area of ​​a basic recognition rectangular region based on the basic recognition parameter information, obtain the image area of ​​a traffic accident scene based on the image size information, and obtain reference object information of the accident scene based on the image information of the traffic accident scene.

[0088] Optionally, the evaluation module specifically includes:

[0089] a first evaluation unit, configured to use a ratio of an area of ​​a basic identification rectangular region to an area of ​​a traffic accident scene image as an image recognition coefficient, obtain physical size information corresponding to an accident scene reference object, use a ratio of an image size to a physical size as an image scale conversion coefficient, and obtain a basic accident scene reference object and a reference object calibration group based on the image recognition coefficient;

[0090] The second evaluation unit is configured to obtain a first range curve and a second range curve corresponding to the first reference object and the second reference object on site, and obtain image mapping points according to the first range curve and the second range curve.

[0091] Compared with the prior art, the present invention has the following beneficial effects:

[0092] The present invention proposes a traffic accident scene analysis method and system based on multimodal data. The method determines the image recognition status through the image recognition coefficient, which provides a basis for subsequent image recognition settings. The method performs image recognition of different specifications on images of different areas through image region classification information, thereby ensuring the accuracy of image recognition while improving the image recognition efficiency. The method reversely infers the location of the traffic accident through the accident target information, realizes the analysis of the traffic accident intensity, and improves the analysis efficiency and accuracy of the traffic accident scene. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 This is a flow chart of a traffic accident scene analysis method based on multimodal data proposed by the present invention;

[0094] Figure 2 This is a flowchart for obtaining accident target information in the present invention;

[0095] Figure 3 A flow chart for obtaining a reference calibration group in the present invention;

[0096] Figure 4 This is a flowchart of image acquisition height acquisition in the present invention;

[0097] Figure 5 This is a structural block diagram of a traffic accident scene analysis system based on multimodal data proposed by the present invention. DETAILED DESCRIPTION

[0098] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0099] Reference Figure 1 - Figure 4 As shown, a traffic accident scene analysis method based on multimodal data in an embodiment of the present invention includes:

[0100] Acquiring traffic accident scene information, wherein the traffic accident scene information includes traffic accident scene image information and scene map data;

[0101] Based on the traffic accident scene image information, image parameter information is obtained, wherein the image parameter information includes image resolution information and image size information;

[0102] According to the traffic accident scene information, target recognition is performed based on the YOLO algorithm to obtain accident target information;

[0103] Specifically, based on the traffic accident scene information, the target is identified based on the YOLO algorithm to obtain the accident target information, including:

[0104] According to the image parameter information, obtain image resolution information and image size information;

[0105] Based on the image resolution information, basic recognition parameter information is obtained, where the basic recognition parameter information represents the minimum length and minimum width of a recognizable object;

[0106] According to the basic recognition parameter information, the area of ​​the basic recognition rectangular region is obtained;

[0107] Obtain the image area of ​​the traffic accident scene according to the image size information;

[0108] The ratio of the area of ​​the basic recognition rectangle to the area of ​​the traffic accident scene image is used as the image recognition coefficient;

[0109] Obtaining a basic accident scene reference object and a reference object calibration group according to the image recognition coefficient, wherein the reference object calibration group includes a first reference object at the scene and a second reference object at the scene;

[0110] According to the reference object calibration group, the image acquisition height is obtained;

[0111] Obtain image area classification information based on basic accident scene reference objects and image acquisition height;

[0112] According to the image area classification information, the accident target is identified based on the YOLO algorithm to obtain accident target information. The accident target information includes a first accident target and a second accident target. The first accident target represents the main target of the traffic accident, and the second accident target represents the scattered objects after the traffic accident.

[0113] In this solution, image parameters are first used to obtain resolution and size information, and basic recognition parameters (minimum length and width) and recognition coefficients are derived to define the "capability boundaries" of the YOLO algorithm. This approach clearly defines the basic conditions for target recognition, avoiding recognition bias due to differences in image characteristics. The algorithm is then adapted to the scene imagery, improving initial target recognition accuracy. On-site information such as reference object calibration groups and image acquisition height is then used to associate basic reference objects to complete image region classification. Image data is integrated with spatial and physical information at the scene, addressing the inadequacy of scene understanding in image recognition alone. This allows the YOLO algorithm to identify targets within the classified areas, more accurately matching the actual accident scene and achieving accurate identification of accident targets.

[0114] It is understandable that when using the YOLO algorithm to identify targets in an image, it is necessary to first segment the image according to a certain number of grids, and then analyze the image within each grid and merge the grids corresponding to the target to achieve target identification. Therefore, the number of grids is closely related to the recognition accuracy. For images of traffic accident scenes, it is necessary not only to identify the main body of the traffic accident (such as vehicles, pedestrians, etc.), but also to identify the scattered objects generated when the accident occurs. If the number of grids is small and the recognition accuracy is low, the scattered objects cannot be accurately identified. However, as the number of grids increases, the amount of data output also gradually increases, which not only increases the data processing resource usage but also reduces the image recognition efficiency. Therefore, in this solution, different division criteria are selected for different areas to ensure image recognition accuracy while improving image recognition efficiency.

[0115] It should be noted that, in this embodiment, the original traffic accident scene image is divided according to the grid number standards of 104×104, 52×52, and 26×26, respectively, and the grid width setting corresponding to each grid number standard is obtained. If it is the first image area, the area is divided according to the grid number standard of 26×26; if it is the second image area, the area is divided according to the grid number standard of 52×52; if it is the third image area, the area is divided according to the grid number standard of 104×104.

[0116] Specifically, according to the image recognition coefficient, the basic accident scene reference object and the reference object calibration group are obtained, which specifically includes:

[0117] Acquiring accident scene reference object information based on traffic accident scene image information, wherein the accident scene reference object information includes image size information corresponding to each accident scene reference object, and the accident scene reference objects include lane lines, guardrails, and traffic signs;

[0118] Obtain the physical size information corresponding to the reference object at the accident scene, and use the ratio of the image size to the physical size as the image scale conversion coefficient;

[0119] The accident scene reference object corresponding to the maximum value of the image scale conversion coefficient is used as the first reference object at the scene;

[0120] The product of the maximum value of the image scale conversion coefficient and the image recognition coefficient is used as the image recognition coefficient threshold;

[0121] Screening accident scene reference objects according to an image recognition coefficient threshold to obtain basic accident scene reference objects;

[0122] If the image scale conversion coefficient corresponding to the accident scene reference object does not exceed the image recognition coefficient threshold, the accident scene reference object is removed;

[0123] The basic accident scene reference object corresponding to the minimum value of the image scale conversion coefficient among the basic accident scene reference objects is used as the second reference object at the scene;

[0124] A reference object calibration group is obtained according to the first on-site reference object and the second on-site reference object.

[0125] This approach focuses on the selection and calibration of reference objects, laying a solid foundation for subsequent analysis. Starting with reference objects at the accident scene (such as lane markings), a threshold is established using the ratio of image to physical size (scaling coefficient) combined with the image recognition coefficient. Reference objects with scale conversion coefficients that do not exceed the threshold are selected as baseline reference objects, while those with poor compatibility are eliminated. This ensures that the reference objects involved in the analysis are compatible with the image recognition capabilities, providing a reliable "anchor" for subsequent reference-based analysis (such as acquisition height calculation). A calibration group is determined, consisting of the first reference object at the scene (with the largest scale conversion coefficient) and the second reference object (with the smallest coefficient among the baseline reference objects), forming an "extreme value" association for the reference objects. This approach strengthens the mapping relationship between the reference objects in the image and physical space. Subsequent calculations of information such as image acquisition height based on the calibration group can leverage extreme value differences for precise derivation, improving the accuracy of fusion analysis of multimodal data (image and physical scene). As key elements at the scene, their precise selection and calibration facilitate subsequent image region segmentation and accident target identification. By providing scene priors (such as spatial information associated with reference objects) for YOLO algorithm recognition, target recognition can be more closely aligned with the physical logic of the accident scene, promoting the transition from image recognition to in-depth scene understanding and improving the comprehensiveness and scientific nature of accident scene analysis.

[0126] Specifically, according to the reference object calibration group, the image acquisition height is obtained, which specifically includes:

[0127] According to the on-site map data, the first on-site reference object and the second on-site reference object are matched with the on-site map to obtain the position information of the calibration reference objects;

[0128] Acquire a reference object characteristic distance according to the calibration reference object position information, wherein the reference object characteristic distance represents a straight-line distance between a first reference object on site and a second reference object on site;

[0129] Obtaining a reference object conversion distance based on the traffic accident scene image information, wherein the reference object conversion distance represents a straight-line distance between two accident scene reference objects in a reference object calibration group in the traffic accident scene image;

[0130] Obtaining pixel size information and focal length information based on the traffic accident scene image information, wherein the pixel size represents the physical width corresponding to a single pixel;

[0131] The product of the reference object characteristic distance and the focal length is used as the spatial characteristic coefficient, and the product of the reference object conversion distance and the pixel size is used as the plane characteristic coefficient;

[0132] The ratio of the spatial characteristic coefficient to the plane characteristic coefficient is used as the image acquisition height.

[0133] In this solution, the coordinates of the on-site map data and the image data are aligned, the physical distance of the reference object (feature distance) is associated with the image pixel distance (conversion distance), and the pixel size and focal length information are combined to establish a complete "physical space-image plane" mapping relationship. This provides a basic scale benchmark for subsequent three-dimensional scene applications such as accident reconstruction and collision trajectory analysis, upgrades two-dimensional image analysis to three-dimensional scene perception, significantly improves target recognition accuracy, enhances multimodal data fusion capabilities, and provides core technical support for rapid response to traffic accidents, responsibility determination, and prevention research.

[0134] Specifically, image area classification information is obtained based on the basic accident scene reference objects and image acquisition height, including:

[0135] The ratio of focal length to pixel size information is used as the image feature coefficient;

[0136] Obtaining a corresponding image scale conversion coefficient according to a first reference object and a second reference object on site;

[0137] The ratio of the image characteristic coefficient to the image scale conversion coefficient corresponding to the first reference object on the scene is used as the first straight-line distance, and the ratio of the image characteristic coefficient to the image scale conversion coefficient corresponding to the second reference object on the scene is used as the second straight-line distance;

[0138] According to the first straight-line distance and the image acquisition height, the first plane distance is obtained based on the Pythagorean theorem with the first straight-line distance as the hypotenuse of the triangle;

[0139] According to the second straight-line distance and the image acquisition height, the second plane distance is obtained based on the Pythagorean theorem with the second straight-line distance as the hypotenuse of the triangle;

[0140] Taking the first reference object on site as the center and the first plane distance as the radius, a first range curve is obtained;

[0141] Taking the second reference object on site as the center and the second plane distance as the radius, a second range curve is obtained;

[0142] The intersection of the first range curve and the second range curve is used as an image acquisition feature point;

[0143] Based on the basic accident scene reference objects, any one of the basic accident scene reference objects is used as the third reference object on site;

[0144] obtaining a third range curve according to a third reference object on site and an image characteristic coefficient;

[0145] The image acquisition feature point with the smallest distance from the third range curve is used as the image mapping point;

[0146] The distance between the image mapping point and the basic accident scene reference object is used as the image conversion reference value;

[0147] According to the image conversion reference value and the image scale conversion coefficient, the data is fitted with the image conversion reference value as the independent variable and the image scale conversion coefficient as the dependent variable to obtain the image scale characteristic coefficient;

[0148] Based on the image scale characteristic coefficient, an image scale conversion coefficient function is constructed;

[0149] Based on traffic accident target analysis, the minimum size and maximum size of the accident target standard are obtained;

[0150] Obtaining a first distance and a second distance according to a minimum size of an accident target standard, a maximum size of an accident target standard, and an image scale conversion coefficient function;

[0151] Divide the image by taking the image mapping point as the center and the first distance and the second distance as the region range to obtain image region classification information, wherein the image region classification information includes a first image region, a second image region, and a third image region;

[0152] The first distance and the second distance are specifically:

[0153]

[0154] Where, is the first distance, is the second distance, represents the image scale conversion coefficient corresponding to the first distance from the image mapping point, represents the image scale conversion coefficient corresponding to the second distance from the image mapping point, Indicates the standard minimum size of the accident target, Indicates the maximum size of the accident target standard, is the image scale conversion coefficient function, is the image scale characteristic coefficient, Indicates the image conversion reference value, Indicates the image acquisition height.

[0155] This solution provides a precise spatial framework for traffic accident scene analysis by constructing an image scale conversion model, geometric positioning, and region division. Combining image acquisition height and reference object information, it derives planar distance and range curves, locates image acquisition feature points and mapping points, and constructs an image scale conversion coefficient function. By correlating image pixel information with physical spatial depth, it accurately characterizes the image scale conversion patterns at different distances, providing a standardized spatial model for target recognition and size determination, addressing size misjudgments caused by image perspective distortion. Based on the standard size and scale function of accident targets, the primary, secondary, and tertiary image regions are divided. Regions are differentiated according to target characteristics, allowing the YOLO algorithm to focus on the appropriate region for recognition, avoiding interference between different target types and optimizing scene understanding through multimodal data fusion.

[0156] In this embodiment, with the image mapping point as the center, the area within a first distance from the image mapping point is used as the first image area, the area within a second distance from the image mapping point is used as the second image area, and the area outside the second distance from the image mapping point is used as the third image area.

[0157] Obtaining first accident target information and second accident target information according to the accident target information;

[0158] Acquiring characteristic trace information corresponding to a first accident target based on traffic accident scene image information, wherein the characteristic trace information represents brake mark information or skid mark information;

[0159] According to the first accident target, based on the characteristic traces, the accident target baseline is obtained;

[0160] According to the accident target baseline and the second accident target information, the traffic accident intensity is analyzed to obtain traffic accident information.

[0161] Specifically, based on the accident target baseline and the second accident target information, the traffic accident intensity is analyzed to obtain traffic accident information, including:

[0162] Based on the analysis of traffic accident traces, the maximum deviation angle can be obtained by reverse estimating the traces;

[0163] According to the accident target baseline corresponding to the first accident target, the intersection of the accident target baseline is used as the basic accident occurrence point;

[0164] Taking the first accident target as the starting point, the accident target baseline is offset along both sides with half of the maximum deviation angle of the trace reverse estimation as the adjustment range, and the area formed by the basic accident occurrence point is regarded as the accident occurrence area;

[0165] obtaining second accident target density information according to the second accident target information;

[0166] The distance between each second accident target and the image mapping point is used as the image offset coefficient of the second accident target;

[0167] Obtaining an image scale conversion coefficient corresponding to each second accident target according to the image offset coefficient and the image scale conversion coefficient function;

[0168] The square of the image scale conversion coefficient is used as the image density correction coefficient corresponding to each second accident target;

[0169] The product of the second accident target density and the image density correction coefficient is used as the second accident target correction density;

[0170] The second accident target corresponding to the maximum value of the second accident target correction density is used as the accident base point;

[0171] The point closest to the accident base point in the accident area is taken as the accident occurrence point;

[0172] The sum of the distances between the first accident target and the accident occurrence point is used as the first accident intensity index;

[0173] acquiring a second accident target area according to the second accident target information;

[0174] The ratio of the area corresponding to each second accident target correction density to the second accident target area is used as the weight of the second accident target correction density;

[0175] Obtaining a second accident intensity index based on a weighted summation of the second accident target correction density and the corresponding weight;

[0176] The sum of the first accident intensity index and the second accident intensity index is used as the traffic accident intensity index;

[0177] Traffic accident information is obtained based on the first accident target information, the second accident target information, the accident occurrence point and the traffic accident intensity index.

[0178] This solution achieves quantitative assessment and precise location of traffic accident intensity through trace analysis, spatial positioning, and density modeling. The accident location is constructed based on the maximum deviation angle of traces, upgrading traditional single-point positioning to probabilistic area estimation, ensuring the accuracy of the accident location inference. By combining the spatial relationship between the primary accident target (the vehicle) and the secondary accident target (the debris), the density correction coefficient and distance weighting are used to cross-validate the debris-dense area with the accident area to avoid positioning bias. Density is corrected using the square of the image scale conversion coefficient to compensate for visual density differences caused by perspective distortion (e.g., debris appears denser at a distance), ensuring that density analysis is physically accurate. Through geometric modeling and data fusion, traffic accident analysis is upgraded from "qualitative description" to "quantitative analysis," achieving breakthroughs in positioning accuracy, intensity assessment, and process reconstruction. Particularly in complex scenarios such as multi-vehicle collisions and high-speed accidents, its multi-dimensional data collaboration and physical model-driven analysis capabilities provide core technical support for rapid traffic accident response, liability determination, and prevention research.

[0179] In this embodiment, the first accident target is taken as the starting point, and the accident target baseline is offset to both sides. The maximum offset angle is half of the maximum deviation angle of the trace reverse deduction. By dynamically adjusting the accident target baseline, the area formed by the intersection of the accident target baseline is used as the accident occurrence area. If the accident target baseline does not intersect, the point with the smallest distance from the accident base point in the area formed by the accident target baseline is used as the accident occurrence point.

[0180] Reference Figure 5 As shown, further, in combination with the above-mentioned traffic accident scene analysis method based on multimodal data, a traffic accident scene analysis system based on multimodal data is proposed, including:

[0181] A main control module, wherein the main control module is used to fit the data according to the image conversion reference value and the image scale conversion coefficient, with the image conversion reference value as the independent variable and the image scale conversion coefficient as the dependent variable, to obtain the image scale feature coefficient, to construct an image scale conversion coefficient function based on the image scale feature coefficient, to obtain the standard minimum size of the accident target and the standard maximum size of the accident target based on the traffic accident target analysis, to obtain the first distance and the second distance based on the standard minimum size of the accident target, the standard maximum size of the accident target and the image scale conversion coefficient function, to divide the image with the image mapping point as the center and the first distance and the second distance as the region range, to obtain image region classification information, to identify the accident target based on the image region classification information based on the YOLO algorithm, to obtain accident target information, and to obtain traffic accident information based on the accident target information;

[0182] an information acquisition module, the information acquisition module being used to acquire traffic accident scene information, the traffic accident scene information including traffic accident scene image information and scene map data, acquiring image parameter information based on the traffic accident scene image information, the image parameter information including image resolution information and image size information, acquiring basic recognition parameter information based on the image resolution information, the basic recognition parameter information indicating the minimum length and minimum width of a recognizable target, acquiring the area of ​​a basic recognition rectangular region based on the basic recognition parameter information, acquiring the image area of ​​the traffic accident scene based on the image size information, and acquiring reference object information at the accident scene based on the traffic accident scene image information;

[0183] An evaluation module, the evaluation module being configured to use the ratio of the area of ​​the basic identification rectangular area to the area of ​​the traffic accident scene image as an image recognition coefficient, obtain physical size information corresponding to the accident scene reference object, use the ratio of the image size to the physical size as an image scale conversion coefficient, obtain the basic accident scene reference object and the reference object calibration group based on the image recognition coefficient, obtain the corresponding first range curve and second range curve based on the first reference object and the second reference object at the scene, and obtain image mapping points based on the first range curve and the second range curve;

[0184] The display module interacts with the main control module and is used to output and display image area classification information, accident target information, accident target baseline, accident occurrence point and traffic accident information.

[0185] Main control module, specifically including:

[0186] A control unit, the control unit being configured to obtain a standard minimum size and a standard maximum size of an accident target based on traffic accident target analysis, obtain a first distance and a second distance based on the standard minimum size and the standard maximum size of the accident target and an image scale conversion coefficient function, divide the image into regions centered on the image mapping point and using the first and second distances as region ranges, obtain image region classification information, identify the accident target based on the YOLO algorithm based on the image region classification information, obtain accident target information, and obtain traffic accident information based on the accident target information;

[0187] An information receiving unit, which interacts with the information acquisition module and the evaluation module to receive data and transmit it to the data processing unit;

[0188] The data processing unit is used to fit the data according to the image conversion reference value and the image scale conversion coefficient, with the image conversion reference value as the independent variable and the image scale conversion coefficient as the dependent variable, to obtain the image scale characteristic coefficient, and to construct an image scale conversion coefficient function based on the image scale characteristic coefficient.

[0189] Information acquisition module, specifically including:

[0190] a first acquisition unit, configured to acquire traffic accident scene information, the traffic accident scene information including traffic accident scene image information and scene map data, and acquire image parameter information based on the traffic accident scene image information, the image parameter information including image resolution information and image size information;

[0191] The second acquisition unit is used to obtain basic recognition parameter information based on image resolution information, where the basic recognition parameter information represents the minimum length and minimum width of an identifiable target, obtain the area of ​​a basic recognition rectangular region based on the basic recognition parameter information, obtain the image area of ​​a traffic accident scene based on the image size information, and obtain reference object information of the accident scene based on the image information of the traffic accident scene.

[0192] Assessment modules include:

[0193] a first evaluation unit, configured to use a ratio of an area of ​​a basic identification rectangular region to an area of ​​a traffic accident scene image as an image recognition coefficient, obtain physical size information corresponding to an accident scene reference object, use a ratio of an image size to a physical size as an image scale conversion coefficient, and obtain a basic accident scene reference object and a reference object calibration group based on the image recognition coefficient;

[0194] The second evaluation unit is configured to obtain a first range curve and a second range curve corresponding to the first reference object and the second reference object on site, and obtain image mapping points according to the first range curve and the second range curve.

[0195] To sum up, the advantages of the present invention are: by taking the ratio of the area of ​​the basic recognition rectangular area and the area of ​​the traffic accident scene image as the image recognition coefficient, determining the image recognition status provides a basis for subsequent image recognition settings, and obtaining image area classification information through the basic accident scene reference object and image acquisition height. Through the image area classification information, image recognition of different specifications is performed on images of different areas, ensuring the accuracy of image recognition while improving the image recognition efficiency. The location of the traffic accident is reversed through the accident target information, and the intensity of the traffic accident is analyzed, thereby improving the analysis efficiency and accuracy of the traffic accident scene.

[0196] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A traffic accident scene analysis method based on multimodal data, characterized in that: include: Acquiring traffic accident scene information, wherein the traffic accident scene information includes traffic accident scene image information and scene map data; Based on the traffic accident scene image information, image parameter information is obtained, wherein the image parameter information includes image resolution information and image size information; According to the traffic accident scene information, target recognition is performed based on the YOLO algorithm to obtain accident target information; Obtaining first accident target information and second accident target information according to the accident target information; Acquiring characteristic trace information corresponding to a first accident target based on traffic accident scene image information, wherein the characteristic trace information represents brake mark information or skid mark information; According to the first accident target, based on the characteristic traces, the accident target baseline is obtained; Analyze the traffic accident intensity based on the accident target baseline and the second accident target information to obtain traffic accident information; The method of identifying the target based on the YOLO algorithm according to the traffic accident scene information to obtain the accident target information specifically includes: According to the image parameter information, obtain image resolution information and image size information; Based on the image resolution information, basic recognition parameter information is obtained, where the basic recognition parameter information represents the minimum length and minimum width of a recognizable object; According to the basic recognition parameter information, the area of ​​the basic recognition rectangular region is obtained; Obtain the image area of ​​the traffic accident scene according to the image size information; The ratio of the area of ​​the basic recognition rectangle to the area of ​​the traffic accident scene image is used as the image recognition coefficient; Obtaining a basic accident scene reference object and a reference object calibration group according to the image recognition coefficient, wherein the reference object calibration group includes a first reference object at the scene and a second reference object at the scene; According to the reference object calibration group, the image acquisition height is obtained; Obtain image area classification information based on basic accident scene reference objects and image acquisition height; According to the image area classification information, the accident target is identified based on the YOLO algorithm to obtain accident target information. The accident target information includes a first accident target and a second accident target. The first accident target represents the main target of the traffic accident, and the second accident target represents the scattered objects after the traffic accident.

2. The traffic accident scene analysis method based on multimodal data according to claim 1, characterized in that: The step of obtaining the basic accident scene reference object and the reference object calibration group according to the image recognition coefficient specifically includes: Acquiring accident scene reference object information based on traffic accident scene image information, wherein the accident scene reference object information includes image size information corresponding to each accident scene reference object, and the accident scene reference objects include lane lines, guardrails, and traffic signs; Obtain the physical size information corresponding to the reference object at the accident scene, and use the ratio of the image size to the physical size as the image scale conversion coefficient; The accident scene reference object corresponding to the maximum value of the image scale conversion coefficient is used as the first reference object at the scene; The product of the maximum value of the image scale conversion coefficient and the image recognition coefficient is used as the image recognition coefficient threshold; Screening accident scene reference objects according to an image recognition coefficient threshold to obtain basic accident scene reference objects; If the image scale conversion coefficient corresponding to the accident scene reference object does not exceed the image recognition coefficient threshold, the accident scene reference object is removed; The basic accident scene reference object corresponding to the minimum value of the image scale conversion coefficient among the basic accident scene reference objects is used as the second reference object at the scene; A reference object calibration group is obtained according to the first on-site reference object and the second on-site reference object.

3. The traffic accident scene analysis method based on multimodal data according to claim 2, characterized in that: The step of obtaining the image acquisition height according to the reference object calibration group specifically includes: According to the on-site map data, the first on-site reference object and the second on-site reference object are matched with the on-site map to obtain the position information of the calibration reference objects; Acquire a reference object characteristic distance according to the calibration reference object position information, wherein the reference object characteristic distance represents a straight-line distance between a first reference object on site and a second reference object on site; Obtaining a reference object conversion distance based on the traffic accident scene image information, wherein the reference object conversion distance represents a straight-line distance between two accident scene reference objects in a reference object calibration group in the traffic accident scene image; Obtaining pixel size information and focal length information based on the traffic accident scene image information, wherein the pixel size represents the physical width corresponding to a single pixel; The product of the reference object characteristic distance and the focal length is used as the spatial characteristic coefficient, and the product of the reference object conversion distance and the pixel size is used as the plane characteristic coefficient; The ratio of the spatial characteristic coefficient to the plane characteristic coefficient is used as the image acquisition height.

4. The traffic accident scene analysis method based on multimodal data according to claim 3, characterized in that: The obtaining of image area classification information based on the basic accident scene reference object and the image acquisition height specifically includes: The ratio of focal length to pixel size information is used as the image feature coefficient; Obtaining a corresponding image scale conversion coefficient according to a first reference object and a second reference object on site; The ratio of the image characteristic coefficient to the image scale conversion coefficient corresponding to the first reference object on the scene is used as the first straight-line distance, and the ratio of the image characteristic coefficient to the image scale conversion coefficient corresponding to the second reference object on the scene is used as the second straight-line distance; According to the first straight-line distance and the image acquisition height, the first plane distance is obtained based on the Pythagorean theorem with the first straight-line distance as the hypotenuse of the triangle; According to the second straight-line distance and the image acquisition height, the second plane distance is obtained based on the Pythagorean theorem with the second straight-line distance as the hypotenuse of the triangle; Taking the first reference object on site as the center and the first plane distance as the radius, a first range curve is obtained; Taking the second reference object on site as the center and the second plane distance as the radius, a second range curve is obtained; The intersection of the first range curve and the second range curve is used as an image acquisition feature point; Based on the basic accident scene reference objects, any one of the basic accident scene reference objects is used as the third reference object on site; obtaining a third range curve according to a third reference object on site and an image characteristic coefficient; The image acquisition feature point with the smallest distance from the third range curve is used as the image mapping point; The distance between the image mapping point and the basic accident scene reference object is used as the image conversion reference value; According to the image conversion reference value and the image scale conversion coefficient, the data is fitted with the image conversion reference value as the independent variable and the image scale conversion coefficient as the dependent variable to obtain the image scale characteristic coefficient; Based on the image scale characteristic coefficient, an image scale conversion coefficient function is constructed; Based on traffic accident target analysis, the minimum size and maximum size of the accident target standard are obtained; Obtaining a first distance and a second distance according to a minimum size of an accident target standard, a maximum size of an accident target standard, and an image scale conversion coefficient function; Divide the image by taking the image mapping point as the center and the first distance and the second distance as the region range to obtain image region classification information, wherein the image region classification information includes a first image region, a second image region, and a third image region; The first distance and the second distance are specifically: Where, is the first distance, is the second distance, represents the image scale conversion coefficient corresponding to the first distance from the image mapping point, represents the image scale conversion coefficient corresponding to the second distance from the image mapping point, Indicates the minimum size of the accident target standard, Indicates the maximum size of the accident target standard, is the image scale conversion coefficient function, is the image scale characteristic coefficient, Indicates the image conversion reference value, Indicates the image acquisition height.

5. The traffic accident scene analysis method based on multimodal data according to claim 4 is characterized in that: The analyzing the traffic accident intensity based on the accident target baseline and the second accident target information to obtain traffic accident information specifically includes: Based on the analysis of traffic accident traces, the maximum deviation angle can be obtained by reverse estimating the traces; According to the accident target baseline corresponding to the first accident target, the intersection of the accident target baseline is used as the basic accident occurrence point; Taking the first accident target as the starting point, the accident target baseline is offset along both sides with half of the maximum deviation angle of the trace reverse estimation as the adjustment range, and the area formed by the basic accident occurrence point is regarded as the accident occurrence area; obtaining second accident target density information according to the second accident target information; The distance between each second accident target and the image mapping point is used as the image offset coefficient of the second accident target; Obtaining an image scale conversion coefficient corresponding to each second accident target according to the image offset coefficient and the image scale conversion coefficient function; The square of the image scale conversion coefficient is used as the image density correction coefficient corresponding to each second accident target; The product of the second accident target density and the image density correction coefficient is used as the second accident target correction density; The second accident target corresponding to the maximum value of the second accident target correction density is used as the accident base point; The point in the accident area closest to the accident base point is taken as the accident occurrence point; The sum of the distances between the first accident target and the accident occurrence point is used as the first accident intensity index; acquiring a second accident target area according to the second accident target information; The ratio of the area corresponding to each second accident target correction density to the second accident target area is used as the weight of the second accident target correction density; According to the second accident target correction density and the corresponding weight, a second accident intensity index is obtained based on weighted summation; The sum of the first accident intensity index and the second accident intensity index is used as the traffic accident intensity index; Traffic accident information is obtained based on the first accident target information, the second accident target information, the accident occurrence point and the traffic accident intensity index.

6. A traffic accident scene analysis system based on multimodal data, used to implement the analysis method according to any one of claims 1 to 5, characterized in that: include: A main control module, wherein the main control module is used to fit the data according to the image conversion reference value and the image scale conversion coefficient, with the image conversion reference value as the independent variable and the image scale conversion coefficient as the dependent variable, to obtain the image scale feature coefficient, to construct an image scale conversion coefficient function based on the image scale feature coefficient, to obtain the standard minimum size of the accident target and the standard maximum size of the accident target based on the traffic accident target analysis, to obtain the first distance and the second distance based on the standard minimum size of the accident target, the standard maximum size of the accident target and the image scale conversion coefficient function, to divide the image with the image mapping point as the center and the first distance and the second distance as the region range, to obtain image region classification information, to identify the accident target based on the image region classification information based on the YOLO algorithm, to obtain accident target information, and to obtain traffic accident information based on the accident target information; an information acquisition module, the information acquisition module being used to acquire traffic accident scene information, the traffic accident scene information including traffic accident scene image information and scene map data, acquiring image parameter information based on the traffic accident scene image information, the image parameter information including image resolution information and image size information, acquiring basic recognition parameter information based on the image resolution information, the basic recognition parameter information indicating the minimum length and minimum width of a recognizable target, acquiring the area of ​​a basic recognition rectangular region based on the basic recognition parameter information, acquiring the image area of ​​the traffic accident scene based on the image size information, and acquiring reference object information at the accident scene based on the traffic accident scene image information; An evaluation module, the evaluation module being configured to use the ratio of the area of ​​the basic identification rectangular area to the area of ​​the traffic accident scene image as an image recognition coefficient, obtain physical size information corresponding to the accident scene reference object, use the ratio of the image size to the physical size as an image scale conversion coefficient, obtain the basic accident scene reference object and the reference object calibration group based on the image recognition coefficient, obtain the corresponding first range curve and second range curve based on the first reference object and the second reference object at the scene, and obtain image mapping points based on the first range curve and the second range curve; The display module interacts with the main control module and is used to output and display image area classification information, accident target information, accident target baseline, accident occurrence point and traffic accident information.

7. The traffic accident scene analysis system based on multimodal data according to claim 6, characterized in that: The main control module specifically includes: A control unit, the control unit being configured to obtain a standard minimum size and a standard maximum size of an accident target based on traffic accident target analysis, obtain a first distance and a second distance based on the standard minimum size and the standard maximum size of the accident target and an image scale conversion coefficient function, divide the image into regions centered on the image mapping point and using the first and second distances as region ranges, obtain image region classification information, identify the accident target based on the YOLO algorithm based on the image region classification information, obtain accident target information, and obtain traffic accident information based on the accident target information; An information receiving unit, which interacts with the information acquisition module and the evaluation module to receive data and transmit it to the data processing unit; The data processing unit is used to fit the data according to the image conversion reference value and the image scale conversion coefficient, with the image conversion reference value as the independent variable and the image scale conversion coefficient as the dependent variable, to obtain the image scale characteristic coefficient, and to construct an image scale conversion coefficient function based on the image scale characteristic coefficient.

8. The traffic accident scene analysis system based on multimodal data according to claim 6, characterized in that: The information acquisition module specifically includes: a first acquisition unit, configured to acquire traffic accident scene information, the traffic accident scene information including traffic accident scene image information and scene map data, and acquire image parameter information based on the traffic accident scene image information, the image parameter information including image resolution information and image size information; The second acquisition unit is used to obtain basic recognition parameter information based on image resolution information, where the basic recognition parameter information represents the minimum length and minimum width of an identifiable target, obtain the area of ​​a basic recognition rectangular region based on the basic recognition parameter information, obtain the image area of ​​a traffic accident scene based on the image size information, and obtain reference object information of the accident scene based on the image information of the traffic accident scene.

9. The traffic accident scene analysis system based on multimodal data according to claim 6, characterized in that: The evaluation module specifically includes: a first evaluation unit, configured to use a ratio of an area of ​​a basic identification rectangular region to an area of ​​a traffic accident scene image as an image recognition coefficient, obtain physical size information corresponding to an accident scene reference object, use a ratio of an image size to a physical size as an image scale conversion coefficient, and obtain a basic accident scene reference object and a reference object calibration group based on the image recognition coefficient; The second evaluation unit is configured to obtain a first range curve and a second range curve corresponding to the first reference object and the second reference object on site, and obtain image mapping points according to the first range curve and the second range curve.

Images