Traffic accident scene reconstruction method, device, equipment and medium

Abstract

The invention discloses a traffic accident scene reconstruction method and device, equipment and a medium, and relates to the technical field of computer vision. Comprising the following steps: acquiring an initial point cloud and a multi-view image of a traffic accident scene, and determining an initial scanning point based on the initial point cloud and the multi-view image; generating a global path and a dynamic partition priority table based on the initial scanning point, and generating a scanning path based on the global path and the dynamic partition priority table; and carrying out area scanning on the traffic accident scene based on the scanning path to obtain image data and point cloud data corresponding to each area of the traffic accident scene, and carrying out fusion processing on the image data and the point cloud data to construct a three-dimensional model corresponding to the traffic accident scene. Therefore, the automation level of three-dimensional reconstruction can be improved, the scene coverage rate and data consistency of reconstruction are improved, efficient processing of complex scenes is realized, and the reconstruction efficiency and quality are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and in particular to a method, device, equipment and medium for reconstructing a traffic accident scene. Background Art

[0002] At present, the 3D reconstruction technology of traffic accident scenes mainly relies on handheld 3D reconstruction equipment. Handheld equipment usually uses laser scanning or multi-view imaging technology. The scanning path and angle are completely dependent on the operator's proficiency. It is difficult for operators to cover some special areas (such as the bottom or high places of vehicles), which may lead to partial data loss. In addition, lighting conditions, dynamic obstacles and dangerous environments on site (such as traffic flow or burning) will seriously affect equipment performance and operational safety.

[0003] The existing handheld 3D reconstruction technology requires personnel to enter the traffic accident scene, which has the disadvantages of high risk, low efficiency, and insufficient coverage, which limits its actual application effect in traffic accident 3D reconstruction. Therefore, how to improve the automation level of 3D reconstruction, improve the reconstruction scene coverage and data consistency, achieve efficient processing of complex scenes, and improve reconstruction efficiency and quality are the problems that need to be solved urgently. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a traffic accident scene reconstruction method, device, equipment and medium, which can improve the automation level of three-dimensional reconstruction, enhance the reconstruction scene coverage and data consistency, realize efficient processing of complex scenes, and improve reconstruction efficiency and quality. The specific scheme is as follows:

[0005] In a first aspect, the present application provides a method for reconstructing a traffic accident scene, which is applied to a wheeled robot, comprising:

[0006] Acquire an initial point cloud and a multi-view image of the traffic accident scene, and determine a starting scanning point based on the initial point cloud and the multi-view image;

[0007] Generate a global path and a dynamic partition priority table based on the starting scanning point, and generate a scanning path based on the global path and the dynamic partition priority table; the dynamic partition priority table is used to characterize the reconstruction priority of each area partition of the traffic accident scene;

[0008] The traffic accident scene is regionally scanned based on the scanning path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene, and the image data and the point cloud data are fused to construct a three-dimensional model corresponding to the traffic accident scene.

[0009] Optionally, the wheeled robot includes a fisheye camera array, a laser radar, a multi-sensor fusion component, a lifting rod system for controlling the height of the fisheye camera and the laser radar through a motor, and a controller terminal for remotely monitoring the operation of the equipment and managing the reconstruction results;

[0010] Accordingly, the acquisition of the initial point cloud and multi-view images of the traffic accident scene includes:

[0011] The initial point cloud and multi-view images of the traffic accident scene are acquired through the fisheye camera array and the laser radar.

[0012] Optionally, generating a scan path based on the global path and the dynamic partition priority table includes:

[0013] Creating a path key node based on the dynamic partition priority table, and generating a scan path based on the path key node and the global path;

[0014] The path key nodes are used to mark areas where the height of the lifting rod system needs to be adjusted or where the reconstruction priority meets a preset priority condition.

[0015] Optionally, performing regional scanning on the traffic accident scene based on the scanning path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene includes:

[0016] Using the multi-sensor fusion component to dynamically adjust the driving path of the wheeled robot based on the scanning path and the on-site conditions of the traffic accident scene;

[0017] The fisheye camera array and the laser radar are used to perform a regional scan of the traffic accident scene based on the driving path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene.

[0018] Optionally, after performing regional scanning on the traffic accident scene based on the scanning path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene, the method further includes:

[0019] monitoring the point cloud density of the point cloud data in real time, and determining a scanning blind area based on the image data and the point cloud density;

[0020] A supplementary scanning path is planned based on the scanning blind area, and the scanning blind area is scanned based on the supplementary scanning path to obtain image data and point cloud data corresponding to the scanning blind area.

[0021] Optionally, fusing the image data and the point cloud data to construct a three-dimensional model corresponding to the traffic accident scene includes:

[0022] Using a clock synchronization mechanism and an external parameter calibration matrix, the time and space alignment of the point cloud data and the image data is achieved;

[0023] Visual information is used to perform detail compensation on sparse areas of the point cloud data, and the texture of the image data is mapped onto the surface of the point cloud data to construct a three-dimensional model corresponding to the traffic accident scene.

[0024] Optionally, after fusing the image data and the point cloud data to construct a three-dimensional model corresponding to the traffic accident scene, the method further includes:

[0025] The three-dimensional model is stored in a preset storage space and uploaded to a cloud platform so as to preview the three-dimensional model corresponding to the traffic accident scene in real time through the cloud platform.

[0026] In a second aspect, the present application provides a traffic accident scene reconstruction device, which is applied to a wheeled robot, comprising:

[0027] A starting scanning point determination module, used to obtain an initial point cloud and a multi-view image of the traffic accident scene, and determine a starting scanning point based on the initial point cloud and the multi-view image;

[0028] A scanning path planning module, used to generate a global path and a dynamic partition priority table based on the starting scanning point, and generate a scanning path based on the global path and the dynamic partition priority table; the dynamic partition priority table is used to characterize the reconstruction priority of each area partition of the traffic accident scene;

[0029] The three-dimensional model generation module is used to perform regional scanning of the traffic accident scene based on the scanning path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene, and fuse the image data and the point cloud data to construct a three-dimensional model corresponding to the traffic accident scene.

[0030] In a third aspect, the present application provides an electronic device, including:

[0031] Memory, used to store computer programs;

[0032] The processor is used to execute the computer program to implement the aforementioned traffic accident scene reconstruction method.

[0033] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned traffic accident scene reconstruction method.

[0034] The present application provides a traffic accident scene method, which first obtains an initial point cloud and a multi-view image of the traffic accident scene, and determines a starting scanning point based on the initial point cloud and the multi-view image; then generates a global path and a dynamic partition priority table based on the starting scanning point; the dynamic partition priority table is used to characterize the reconstruction priority of each area partition of the traffic accident scene; finally, the traffic accident scene is scanned regionally based on the global path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene, and the image data and the point cloud data are fused to construct a three-dimensional model corresponding to the traffic accident scene.

[0035] As can be seen from the above, the present application utilizes the wheeled robot to obtain the initial point cloud and multi-view images of the traffic accident scene to automatically plan the global route for the traffic accident scene without the need for personnel to enter the dangerous area. The image data and the point cloud data are fused and processed to construct a three-dimensional model corresponding to the traffic accident scene, achieving efficient panoramic scanning and ensuring the consistency of reconstructed data and model quality. This can improve the automation level of three-dimensional reconstruction, improve the reconstruction scene coverage and data consistency, achieve efficient processing of complex scenes, and improve reconstruction efficiency and quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0037] Figure 1 A flow chart of a traffic accident scene reconstruction method applied to a wheeled robot disclosed in the present application;

[0038] Figure 2 A schematic diagram of a traffic accident scene reconstruction device applied to a wheeled robot disclosed in the present application;

[0039] Figure 3 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] At present, the three-dimensional reconstruction technology of traffic accident scenes mainly relies on handheld three-dimensional reconstruction equipment. Handheld devices usually use laser scanning or multi-view imaging technology. The scanning path and angle are completely dependent on the proficiency of the operator. It is difficult for the operator to cover some special areas (such as the bottom or high places of the vehicle), which may cause some data to be missing, and lighting conditions, dynamic obstacles and dangerous environments on site (such as traffic flow or burning) will seriously affect the performance of the equipment and operational safety. The existing handheld three-dimensional reconstruction technology requires personnel to enter the scene of the traffic accident, and has the disadvantages of high danger, low efficiency, and insufficient coverage, which limits its actual application effect in the three-dimensional reconstruction of traffic accidents. To this end, the present application provides a traffic accident scene reconstruction solution that can improve the automation level of three-dimensional reconstruction, improve the reconstruction scene coverage and data consistency, achieve efficient processing of complex scenes, and improve reconstruction efficiency and quality.

[0042] See also Figure 1 As shown, the embodiment of the present application discloses a traffic accident scene reconstruction method applied to a wheeled robot, comprising:

[0043] Step S11, obtaining an initial point cloud and a multi-view image of the traffic accident scene, and determining a starting scanning point based on the initial point cloud and the multi-view image.

[0044] In this embodiment, the wheeled robot is used to obtain the initial point cloud and multi-view images of the traffic accident scene. The wheeled robot includes a fisheye camera array, a laser radar, a multi-sensor fusion component, a lifting rod system that controls the height of the fisheye camera and the laser radar through a motor, and a controller terminal for remotely monitoring the operation of the equipment and managing the reconstruction results. Accordingly, in a specific embodiment, the acquisition of the initial point cloud and multi-view images of the traffic accident scene may include: acquiring the initial point cloud and multi-view images of the traffic accident scene through the fisheye camera array and the laser radar.

[0045] Step S12: Generate a global path and a dynamic partition priority table based on the starting scan point, and generate a scan path based on the global path and the dynamic partition priority table

[0046] In this embodiment, a global path and a dynamic partition priority table are generated based on the initial point cloud and scene layout, and key path nodes are marked on the planned path. The key path nodes are used to mark the areas where the lifting rod system needs to adjust the height or the reconstruction priority meets the preset priority conditions.

[0047] It is worth mentioning that the embodiment of the present application realizes efficient navigation of the robot at the accident scene through the path planning algorithm, and also works in coordination with the sensor lifting mechanism. This collaborative optimization not only improves the comprehensiveness and quality of the reconstructed data, but also reduces the phenomenon of missed scans. Specifically, the generation of the scanning path based on the global path and the dynamic partition priority table may include: creating path key nodes based on the dynamic partition priority table, and generating the scanning path based on the path key nodes and the global path. That is, in the path planning, the areas that need to be reconstructed in the accident scene (such as the collision vehicle, brake marks, and road surface details) are set as priority partitions. Through the fisheye camera detection feedback, the reconstruction order of the sub-areas is adjusted, the partition access order in the main path is dynamically adjusted, and the key areas are processed first, and the path planning algorithm dynamically directs the lifting mechanism to adjust the sensor height according to the spatial characteristics of the partition. For example, when scanning the top of the vehicle, the sensor is raised; when scanning low areas (such as the bottom of the vehicle or brake marks), it is lowered to ensure the best viewing angle and resolution.

[0048] It should be noted that based on the sub-area viewpoint calculated by path planning, the height h of the sensor is dynamically adjusted to cover the entire field of view, where the lifting height optimization formula is as follows:

[0049] ;

[0050] Among them, d(P,S) is the distance between the sensor and the target point, FOV is the field of view of the sensor, and dA is the differential area on the field of view. Through this optimization formula, the sensor can always be kept in the optimal scanning position. It should be noted that the path planning algorithm binds the current position of the robot to the action instructions of the lifting mechanism. For example, when the planned path needs to enter a low area or approach the edge of the vehicle, the lifting mechanism is automatically triggered to adjust to the optimal height to avoid occlusion or blind spots.

[0051] Step S13: performing a regional scan of the traffic accident scene based on the scanning path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene, and fusing the image data and the point cloud data to construct a three-dimensional model corresponding to the traffic accident scene.

[0052] In this embodiment, high-confidence feature points in the global scene are extracted through multi-sensor fusion (fisheye camera and laser radar) to avoid missing key information due to the limitation of the single sensor viewing angle. Specifically, the regional scanning of the traffic accident scene based on the scanning path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene may include: using the multi-sensor fusion component to dynamically adjust the driving path of the wheeled robot based on the scanning path and the on-site conditions of the traffic accident scene; using the fisheye camera array and the laser radar to regionally scan the traffic accident scene based on the driving path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene. It should be noted that edge features and clear texture areas are preferentially selected as feature point candidates in the image data, and points with obvious geometric structure changes, such as surface boundaries and corner points, are preferentially selected in the point cloud data. Among them, the system dynamically adjusts the weight of feature points based on motion estimation and environmental changes, removes redundant feature points through iterative optimization, and improves the uniformity of feature point distribution to avoid excessive concentration in certain areas.

[0053] It is worth mentioning that the path planning algorithm detects blind spots and missed scan areas in reconstruction through point cloud data, and re-divides sub-areas and generates new re-scanning paths. Specifically, after performing a regional scan of the traffic accident scene based on the scanning path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene, it may also include: real-time monitoring of the point cloud density of the point cloud data, and determining the scanning blind area based on the image data and the point cloud density; planning a re-scanning path based on the scanning blind area, and scanning the scanning blind area based on the re-scanning path to obtain image data and point cloud data corresponding to the scanning blind area. Among them, the sparsity function of the point cloud density is as follows:

[0054] ;

[0055] in, Yes The weight of (calculated according to the number of neighborhood points), V is the region volume, and ρ represents the region sparsity.

[0056] It should be noted that the system generates a fill-in scan path based on the sparse area and dynamically adjusts the sensor height during the fill-in scan process. The height adjustment function is as follows:

[0057] ;

[0058] in, is the current sensor height, and are the minimum and maximum height limits, is the adjustment coefficient, is the sparseness threshold. That is, when the density of sparse areas is low, the sensor automatically rises to cover a larger range; when dense areas are identified, the sensor lowers to scan detailed areas.

[0059] In this embodiment, the global features are divided into three levels: points, lines, and surfaces. Different matching strategies are adopted according to different feature dimensions. Combined with the dynamic feature extraction results, the system is refined in the local matching stage, and global consistency constraints (such as loop detection or scene constraints) are used to reduce drift errors and improve the overall structural accuracy of the model. Specifically, the fusion processing of the image data and the point cloud data to construct the three-dimensional model corresponding to the traffic accident scene may include: using a clock synchronization mechanism and an external parameter calibration matrix to achieve time and space alignment of the point cloud data and the image data; using visual information to perform detail compensation on the sparse areas of the point cloud data, and mapping the texture of the image data to the surface of the point cloud data to construct the three-dimensional model corresponding to the traffic accident scene. That is, using a precise clock synchronization mechanism and an external parameter calibration matrix to achieve time and space alignment of the laser radar point cloud and the fisheye camera image data to ensure the accuracy of data fusion; using visual information to perform detail compensation on the sparse areas of the laser radar point cloud, and mapping the high-resolution image texture obtained by the fisheye camera to the point cloud surface, generating a more realistic and delicate three-dimensional model. After the LiDAR point cloud is fused with the image data, the geometric characteristics of the scene (such as ground flatness and complexity) are calculated, and an incremental mapping method is used to locally optimize the existing model after each data collection to ensure that the model is gradually improved during the generation process and reduce the burden of post-processing. The characteristic evaluation function used to calculate the geometric characteristics of the scene is as follows:

[0060] ;

[0061] in, and is the normal vector of the adjacent point, N is the number of adjacent point pairs in the point cloud, and C represents the geometric characteristics of the scene. If the value of C tends to 1, it indicates a flat surface, and if the value of C tends to 0, it indicates a complex structure. It should be noted that the system needs to dynamically adjust the scanning speed and the height of the lifting mechanism according to the value of C.

[0062] Furthermore, the motion control parameters are updated in real time according to the sensor data, wherein the adjustment function for updating the motion control parameters is as follows:

[0063] ;

[0064] in, is the robot's moving speed, and To adjust the parameters, is the threshold of feature complexity; that is, when the scene complexity is high, the speed is slowed down to improve scanning accuracy.

[0065] In this embodiment, relevant personnel can monitor the scanning status and preview the reconstructed color point cloud map of the accident scene in real time through the control terminal. Specifically, after the image data and the point cloud data are fused to construct the three-dimensional model corresponding to the traffic accident scene, the three-dimensional model can also be stored in a preset storage space and uploaded to a cloud platform, so as to preview the reconstructed color point cloud map of the accident scene corresponding to the traffic accident scene in real time through the cloud platform.

[0066] As can be seen from the above, the embodiment of the present application utilizes the wheeled robot to obtain the initial point cloud and multi-view images of the traffic accident scene to automatically plan the global route for the traffic accident scene, without the need for personnel to enter the dangerous area. The system operates automatically throughout the process, and there is no need to manually adjust the path or sensor position, which greatly reduces the complexity and time cost of operation. The collaborative work of the path planning algorithm and the lifting mechanism ensures seamless scanning from the ground to a high altitude, avoiding the blind spot problem caused by the viewing angle limitation in the traditional method. The image data and the point cloud data are fused and processed to construct a three-dimensional model corresponding to the traffic accident scene, achieving efficient panoramic scanning, and ensuring the consistency of the reconstructed data and the quality of the model. Dynamically adjusting the sensor height ensures that the scanning area is always in the optimal field of view, the reconstructed data has a higher resolution, and the details are captured more richly. Thereby, the automation level of three-dimensional reconstruction can be improved, the reconstruction scene coverage and data consistency can be improved, efficient processing of complex scenes can be achieved, and the reconstruction efficiency and quality can be improved.

[0067] The embodiment of the present application provides a specific method for reconstructing a traffic accident scene applied to a wheeled robot, comprising:

[0068] The embodiments of the present application include a wheeled robot body as a mobile platform for the equipment, equipped with a control module, a power module and a sensor lifting system; a fisheye camera array with a panoramic vision system for capturing all-round image data of the accident scene; a laser radar for generating high-precision point cloud maps and realizing scene structure reconstruction; a lifting rod system for controlling the height of the fisheye camera and the laser radar through a motor to flexibly adapt to the requirements of different scenes; a multi-sensor fusion module that combines differential GPS, wheel speed meter and IMU to achieve high-precision positioning and path planning; a path planning module that uses a dynamic multi-level partitioning algorithm to plan the optimal path to cover all areas of the accident scene; and a control terminal for remotely monitoring equipment operation and managing reconstruction results.

[0069] In this embodiment, after the robot is started, the user can draw the approximate area of ​​the accident on the map on the control terminal. The robot will go to the area for automatic reconstruction. After arriving at the accident scene, the sensor module is first activated to obtain the initial point cloud and multi-view images of the accident scene, and the starting scanning point is automatically identified and calibrated. The system generates a global path and a dynamic partition priority table based on the initial point cloud and scene layout, and plans to mark key nodes on the path (areas that need to adjust the height or pay special attention). The lifting mechanism automatically adjusts the sensor height to the initial position according to the path planning. Then the robot starts to move, moves along the planned path and gradually completes the sub-area scanning. The system avoids obstacles in real time and dynamically adjusts the driving path according to the on-site situation. The sensor scans the scene step by step according to the planned path. The lifting mechanism automatically adjusts according to the change of scene height to ensure that the scanning range is fully covered. At the same time, the system monitors the quality of point cloud and image data in real time, automatically detects blind spots or missed scanning areas, and replans the supplementary scanning path and adjusts the lifting mechanism to complete the supplementary scanning. After completing all partition scans, the system fuses the point cloud and image data, builds a three-dimensional model of the accident scene, and calibrates the overall accuracy. Finally, the system stores the 3D reconstructed model in the device memory and automatically uploads it to the cloud platform. It should be noted that the user can monitor the scanning status in real time and preview the reconstructed color point cloud map of the accident scene through the control terminal.

[0070] The path planning algorithm of the embodiment of the present application realizes efficient navigation of the robot at the accident scene, and can work in conjunction with the sensor lifting mechanism. This collaborative optimization not only improves the comprehensiveness and quality of the reconstructed data, but also reduces the phenomenon of missed scans. Among them, the planning level is divided into two layers: global and local. The global layer quickly generates a main path covering the entire scene according to the approximate GPS area defined by the user on the control terminal, with the shortest time and efficient coverage as the goal. The local layer updates the path in real time according to the sensor data, and repairs dynamic obstacles and missed scan areas. In the path planning, the areas that need to be reconstructed at the accident scene (such as collision vehicles, brake marks, and road surface details) are set as priority partitions. Through the fisheye camera detection feedback, the reconstruction order of the sub-areas is adjusted, and the partition access order in the main path is dynamically adjusted to give priority to the key areas. The path planning algorithm dynamically commands the lifting mechanism to adjust the sensor height according to the spatial characteristics of the partition. For example, when scanning the top of the vehicle, the sensor is raised; when scanning low areas (such as the bottom of the vehicle or brake marks), it is lowered to ensure the best viewing angle and resolution. The path planning algorithm binds the current position of the robot to the action instructions of the lifting mechanism. For example, when the planned path needs to enter a low area or approach the edge of the vehicle, the lifting mechanism is automatically triggered to adjust to the optimal height to avoid occlusion or blind spots. The path planning algorithm detects blind spots and missed scan areas in the reconstruction through point cloud data, re-divides sub-areas and generates new paths. During operation, the system calculates the density of sparse areas in the point cloud based on the 3D model constructed in real time.

[0071] In this embodiment, during the three-dimensional reconstruction process, high-confidence feature points in the global scene are extracted through multi-sensor fusion (fisheye camera and laser radar) to avoid missing key information due to the limitation of the single sensor's viewing angle. The system dynamically adjusts the weight of feature points based on motion estimation and environmental changes, removes redundant feature points through iterative optimization, and improves the uniformity of feature point distribution to avoid excessive concentration in certain areas. Combined with the results of dynamic feature extraction, the system refines in the local matching stage, and uses global consistency constraints (such as loop detection or scene constraints) to reduce drift errors and improve the overall structural accuracy of the model. Using a precise clock synchronization mechanism and an external parameter calibration matrix, the time and space alignment of the laser radar point cloud and the fisheye camera image data is achieved to ensure the accuracy of data fusion. Visual information is used to compensate for details in the sparse areas of the laser radar point cloud, and the high-resolution image texture obtained by the fisheye camera is mapped to the point cloud surface to generate a more realistic and delicate three-dimensional model. Through point cloud density analysis and scene repeated feature detection, redundant scanning data is eliminated to improve reconstruction efficiency and model compactness.

[0072] As can be seen from the above, the embodiment of the present application ensures seamless scanning from the ground to a high place through the coordinated work of the path planning algorithm and the lifting mechanism, avoiding the blind spot problem caused by the viewing angle limitation in the traditional method. Dynamic adjustment of the sensor height ensures that the scanning area is always in the optimal field of view, and the reconstructed data has a higher resolution and richer detail capture. The system runs automatically throughout the process, and there is no need to manually adjust the path or sensor position, which greatly reduces the complexity and time cost of operation. It can improve the automation level of three-dimensional reconstruction, improve the reconstruction scene coverage and data consistency, realize efficient processing of complex scenes, and improve reconstruction efficiency and quality.

[0073] See also Figure 2 As shown, the embodiment of the present application discloses a traffic accident scene reconstruction device applied to a wheeled robot, comprising:

[0074] A starting scanning point determination module 11 is used to obtain an initial point cloud and a multi-view image of the traffic accident scene, and determine a starting scanning point based on the initial point cloud and the multi-view image;

[0075] A scanning path planning module 12, used to generate a global path and a dynamic partition priority table based on the starting scanning point, and to generate a scanning path based on the global path and the dynamic partition priority table; the dynamic partition priority table is used to characterize the reconstruction priority of each area partition of the traffic accident scene;

[0076] The three-dimensional model generation module 13 is used to perform regional scanning of the traffic accident scene based on the scanning path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene, and fuse the image data and the point cloud data to construct a three-dimensional model corresponding to the traffic accident scene.

[0077] As can be seen from the above, the embodiment of the present application utilizes the wheeled robot to obtain the initial point cloud and multi-view images of the traffic accident scene to automatically plan the global route for the traffic accident scene without the need for personnel to enter the dangerous area. The image data and the point cloud data are fused and processed to construct a three-dimensional model corresponding to the traffic accident scene, achieving efficient panoramic scanning and ensuring the consistency of reconstructed data and model quality. This can improve the automation level of three-dimensional reconstruction, improve the reconstruction scene coverage and data consistency, achieve efficient processing of complex scenes, and improve reconstruction efficiency and quality.

[0078] In some specific embodiments, the wheeled robot may specifically include: a fisheye camera array, a laser radar, a multi-sensor fusion component, a lifting rod system for controlling the height of the fisheye camera and the laser radar through a motor, and a controller terminal for remotely monitoring the operation of the device and managing the reconstruction results;

[0079] Accordingly, the starting scanning point determination module 11 may specifically include:

[0080] The initial point cloud and multi-view image acquisition unit is used to acquire the initial point cloud and multi-view images of the traffic accident scene through the fisheye camera array and the laser radar.

[0081] Accordingly, the scanning path planning module 12 may specifically include:

[0082] A scan path planning unit, configured to create a path key node based on the dynamic partition priority table, and generate a scan path based on the path key node and the global path;

[0083] The path key nodes are used to mark areas where the height of the lifting rod system needs to be adjusted or where the reconstruction priority meets a preset priority condition.

[0084] Accordingly, the three-dimensional model generating module 13 may specifically include:

[0085] A driving path adjustment unit, used to dynamically adjust the driving path of the wheeled robot based on the scanning path and the on-site conditions of the traffic accident scene by using the multi-sensor fusion component;

[0086] An area scanning unit is used to perform an area scan of the traffic accident scene based on the driving path through the fisheye camera array and the laser radar to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene.

[0087] In some specific implementations, the three-dimensional model generation module 13 may specifically include:

[0088] A point cloud data and image data fusion unit, used to achieve time and space alignment of the point cloud data and the image data by using a clock synchronization mechanism and an external parameter calibration matrix;

[0089] The three-dimensional model generating unit is used to use visual information to perform detail compensation on the sparse area of ​​the point cloud data, and map the texture of the image data to the surface of the point cloud data to construct a three-dimensional model corresponding to the traffic accident scene.

[0090] In some specific embodiments, the traffic accident scene reconstruction device applied to the wheeled robot may further include:

[0091] A point cloud data monitoring unit, used for monitoring the point cloud density of the point cloud data in real time, and determining a scanning blind area based on the image data and the point cloud density;

[0092] A scanning path planning unit, used to plan a scanning path based on the scanning blind area, and scan the scanning blind area based on the scanning path to obtain image data and point cloud data corresponding to the scanning blind area;

[0093] The three-dimensional model storage unit is used to store the three-dimensional model in a preset storage space and upload it to a cloud platform so as to preview the three-dimensional model corresponding to the traffic accident scene in real time through the cloud platform.

[0094] Furthermore, the present application also discloses an electronic device. Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment, and the content in the figure cannot be regarded as any limitation on the scope of use of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the traffic accident scene reconstruction method disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0095] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0096] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0097] The operating system 221 is used to manage and control the hardware devices and computer programs 222 on the electronic device 20, and can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the traffic accident scene reconstruction method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks.

[0098] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein the computer program, when executed by a processor, implements the aforementioned disclosed method for reconstructing a traffic accident scene. The specific steps of the method may refer to the corresponding contents disclosed in the aforementioned embodiments, and will not be described in detail here.

[0099] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0100] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0101] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0102] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0103] The technical solution provided by the present application is introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technicians in this field, according to the idea of ​​the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for reconstructing a traffic accident scene, characterized in that: Applied to wheeled robots, including: Acquire an initial point cloud and a multi-view image of the traffic accident scene, and determine a starting scanning point based on the initial point cloud and the multi-view image; Generate a global path and a dynamic partition priority table based on the starting scanning point, and generate a scanning path based on the global path and the dynamic partition priority table; the dynamic partition priority table is used to characterize the reconstruction priority of each area partition of the traffic accident scene; The traffic accident scene is regionally scanned based on the scanning path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene, and the image data and the point cloud data are fused to construct a three-dimensional model corresponding to the traffic accident scene.

2. The traffic accident scene reconstruction method according to claim 1, characterized in that: The wheeled robot includes a fisheye camera array, a laser radar, a multi-sensor fusion component, a lifting rod system for controlling the height of the fisheye camera and the laser radar through a motor, and a controller terminal for remotely monitoring the operation of the device and managing the reconstruction results; Accordingly, the acquisition of the initial point cloud and multi-view images of the traffic accident scene includes: The initial point cloud and multi-view images of the traffic accident scene are acquired through the fisheye camera array and the laser radar.

3. The traffic accident scene reconstruction method according to claim 2, characterized in that: The generating a scan path based on the global path and the dynamic partition priority table includes: Creating a path key node based on the dynamic partition priority table, and generating a scan path based on the path key node and the global path; The path key nodes are used to mark areas where the height of the lifting rod system needs to be adjusted or where the reconstruction priority meets a preset priority condition.

4. The traffic accident scene reconstruction method according to claim 2, characterized in that: The performing of regional scanning on the traffic accident scene based on the scanning path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene includes: Using the multi-sensor fusion component to dynamically adjust the driving path of the wheeled robot based on the scanning path and the on-site conditions of the traffic accident scene; The fisheye camera array and the laser radar are used to perform a regional scan of the traffic accident scene based on the driving path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene.

5. The traffic accident scene reconstruction method according to claim 1, characterized in that: After performing regional scanning on the traffic accident scene based on the scanning path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene, the method further includes: monitoring the point cloud density of the point cloud data in real time, and determining a scanning blind area based on the image data and the point cloud density; A supplementary scanning path is planned based on the scanning blind area, and the scanning blind area is scanned based on the supplementary scanning path to obtain image data and point cloud data corresponding to the scanning blind area.

6. The traffic accident scene reconstruction method according to claim 1, characterized in that: The fusing the image data and the point cloud data to construct a three-dimensional model corresponding to the traffic accident scene includes: Using a clock synchronization mechanism and an external parameter calibration matrix, the time and space alignment of the point cloud data and the image data is achieved; Visual information is used to perform detail compensation on sparse areas of the point cloud data, and the texture of the image data is mapped onto the surface of the point cloud data to construct a three-dimensional model corresponding to the traffic accident scene.

7. The method for reconstructing a traffic accident scene according to any one of claims 1 to 6, characterized in that: After fusing the image data and the point cloud data to construct a three-dimensional model corresponding to the traffic accident scene, the method further includes: The three-dimensional model is stored in a preset storage space and uploaded to a cloud platform so as to preview the three-dimensional model corresponding to the traffic accident scene in real time through the cloud platform.

8. A traffic accident scene reconstruction device, characterized in that: Applied to wheeled robots, including: A starting scanning point determination module, used to obtain an initial point cloud and a multi-view image of the traffic accident scene, and determine a starting scanning point based on the initial point cloud and the multi-view image; A scanning path planning module, used to generate a global path and a dynamic partition priority table based on the starting scanning point, and generate a scanning path based on the global path and the dynamic partition priority table; the dynamic partition priority table is used to characterize the reconstruction priority of each area partition of the traffic accident scene; The three-dimensional model generation module is used to perform regional scanning of the traffic accident scene based on the scanning path to obtain image data and point cloud data corresponding to each area of ​​the traffic accident scene, and fuse the image data and the point cloud data to construct a three-dimensional model corresponding to the traffic accident scene.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the traffic accident scene reconstruction method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store a computer program, wherein when the computer program is executed by a processor, the method for reconstructing a traffic accident scene as described in any one of claims 1 to 7 is implemented.

Images