Traffic video abnormity management and control method and system based on three-dimensional live-action map

Abstract

The invention discloses a traffic video abnormity management and control method and system based on a three-dimensional live-action map, and relates to the technical field of big data analysis, and the method comprises the steps: building a target road three-dimensional live-action model; based on the traffic monitoring video stream of the target road and the three-dimensional live-action modeling of the target road, generating spatial mapping of the traffic monitoring video stream and the three-dimensional live-action modeling of the target road, and marking abnormal event information of a dynamic object in the target road; performing real-time three-dimensional rendering according to the abnormal event information of the dynamic object in the target road, and analyzing an abnormal risk index of the dynamic object in the target road; and performing three-dimensional live-action modeling visual labeling on the target road based on the abnormal risk index of the dynamic object in the target road, determining an abnormal event type of the dynamic object, and generating a traffic abnormality management and control scheme of the target road. The method has the advantages that the traffic management efficiency is improved, the accident occurrence probability is reduced, and traffic flow management and emergency response are optimized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data analysis, in particular to a traffic video abnormal control method and system based on a three-dimensional real scene map. BACKGROUND

[0002] The traffic video abnormal control of the three-dimensional real scene map refers to real-time monitoring and analysis of traffic flow, road conditions, vehicle behavior, etc. by combining three-dimensional map technology and video monitoring technology, so as to identify and warn of abnormal situations in traffic, and then take appropriate management measures to ensure traffic safety and efficiency.

[0003] The existing traffic video abnormal control mainly relies on Internet of Things data to analyze the traffic state of the target area, but when the traffic flow scene is large, false positives are prone to occur, and passive control can only be performed according to the state of the traffic flow, and cannot predict and intervene in advance based on the traffic behavior of a single target. SUMMARY

[0004] To solve the above technical problems, a traffic video abnormal control method and system based on a three-dimensional real scene map are provided, which solves the problem that the existing traffic video abnormal control mainly relies on Internet of Things data to analyze the traffic state of the target area, but when the traffic flow scene is large, false positives are prone to occur, and passive control can only be performed according to the state of the traffic flow, and cannot predict and intervene in advance based on the traffic behavior of a single target.

[0005] To achieve the above purposes, the technical scheme adopted by the present application is as follows: A traffic video abnormal control method based on a three-dimensional real scene map, comprising: S1, based on a traffic heterogeneous sensor, acquiring traffic monitoring video stream and three-dimensional modeling data of a target road, and establishing a three-dimensional real scene modeling of the target road; the three-dimensional modeling data includes point cloud data, satellite images, and street view data; S2, based on the traffic monitoring video stream of the target road and the three-dimensional real scene modeling of the target road, generating a spatial mapping of the traffic monitoring video stream and the three-dimensional real scene modeling of the target road, and marking abnormal event information of dynamic objects in the target road; S3, real-time three-dimensional rendering according to the abnormal event information of the dynamic objects in the target road, generating a traffic state modeling of the dynamic target road three-dimensional real scene, and analyzing abnormal risk indicators of the dynamic objects in the target road; S4, based on the abnormal risk indicators of the dynamic objects in the target road, visualizing the three-dimensional real scene modeling of the target road, determining the abnormal event type of the dynamic objects, and generating a traffic abnormal control scheme for the target road.

[0006] Preferably, step S1 specifically comprises: Based on the multi-source data of traffic monitoring 3D modeling data of the target road, the ICP iterative nearest point algorithm is used to register the point cloud data to eliminate the bias, and satellite imagery and street view data are used as the texture source for 3D modeling to obtain the initial target road 3D real scene modeling data package. Data preprocessing is performed on the initial 3D reality modeling data package of the target road; Using the GDAL data conversion open-source library, coordinate unification is performed on the traffic monitoring video stream of the target road and the initial 3D real scene modeling data package of the target road to obtain multi-source fused 3D real scene modeling data of the target road; Based on multi-source fusion of target road 3D real scene modeling data, the Poisson surface reconstruction algorithm is used to solve the gradient field of the implicit function, which is consistent with the direction of the normal vector field of the input point cloud data, to generate the target road 3D real scene model. Specifically, the ICP iterative nearest point algorithm is as follows:

[0007] In the formula, The transformation matrix is... To transform the matrix The point obtained by transforming the i-th source point. For the target point cloud There are 1 point, where m is the total amount of point cloud data; The Poisson surface reconstruction algorithm is as follows:

[0008] In the formula, For gradient operators, implicit function At point gradient at, For point The normal vector at that location.

[0009] Preferably, step S2 specifically includes: Based on the traffic monitoring video stream of the target road, according to the 3D real scene model of the target road, several sets of 2D and 3D feature image data point pairs are randomly selected from the traffic monitoring video stream of the target road and substituted into the cv2.solvePnP function to calculate and solve the camera extrinsic parameters, world coordinates and intrinsic parameters of the traffic monitoring video stream of the target road, and perform spatial mapping to obtain the spatial location information of the traffic monitoring video stream of the target road. The traffic monitoring video stream of the target road is divided according to the frame rate of the video stream to obtain the traffic monitoring image frame sequence of the target road.

[0010] Preferably, step S2 further includes: The spatial position information of the traffic monitoring video stream of the target road is taken as an image attribute in the traffic monitoring image frame sequence corresponding to the target road, the Yolov5 algorithm is used to collect the dynamic object motion track in the traffic monitoring image frame sequence of the target road, and the running track feature data of each dynamic object is extracted; the running track feature data includes: speed, acceleration, direction change; The running tracks of the dynamic objects in adjacent frames in the running track feature data of each dynamic object are associated by using the DeepSORT multi-target tracking, the position coordinates of the dynamic objects in the complete data stream are marked, and the traffic monitoring image motion object track frame sequence of the target road is generated.

[0011] Preferably, step S2 further comprises: Based on the Lstm long short-term memory network, a target behavior track anomaly analysis model is established, the traffic monitoring image motion object track frame sequence of the target road is taken as input, the future motion object track is predicted according to the target road traffic monitoring image motion object track history for several track steps, and target road abnormal target behavior label data is generated; Based on the internal and external parameters of the traffic monitoring sensor of the target road, the two-dimensional space coordinates of the target road abnormal target behavior label data are mapped to the three-dimensional real scene modeling of the target road, and the abnormal event information of the dynamic object in the target road is obtained; The target behavior track anomaly analysis model specifically comprises:

[0012] In the formula, is the abnormal event information of the i-th dynamic object in the target road at the j-th unit time, j is the hidden layer state of the i-th dynamic object in the target road at the j-th unit time, t is the track feature data of the i-th dynamic object in the target road at the j-th unit time, is the weight matrix of the i-th dynamic object in the target road, is the bias term of the i-th dynamic object in the target road, t is a sigmoid activation function; An observation window is established according to a unit time, the abnormal event information of the dynamic object in the target road is taken as an observation attribute, and the abnormal behavior feature data of the dynamic object in the target road is extracted; t ​​​​​​​​Based on the abnormal behavior characteristic data of the dynamic objects in the target road, an abnormal behavior risk quantization function is established, and an abnormal risk index of the dynamic objects in the initial target road is calculated; The abnormal behavior risk quantization function is specifically as follows:

[0013] is an abnormal risk index of the jth dynamic object in the target road in the tth unit time, is a motion speed value of the jth dynamic object in the target road in the tth unit time, is an acceleration value of the jth dynamic object in the target road in the tth unit time, is a direction change value of the jth dynamic object in the target road in the tth unit time, the motion speed value mean, the acceleration value mean, and the direction change value mean are sequentially represented, the motion speed weight coefficient, the motion speed weight coefficient, and the direction change weight coefficient are sequentially represented.

[0014] Preferably, the step S3 further comprises: According to the abnormal risk index of the dynamic objects in the initial target road, the interaction effect between the abnormal risk indexes of the independent dynamic objects in the initial target road is verified according to the GAM general additive model, and an interaction causal matrix of the abnormal risk indexes of the dynamic objects in the initial target road is established; According to the covariance analysis, the basic abnormal risk index of each dynamic object in the interaction causal matrix of the abnormal risk indexes of the dynamic objects in the initial target road and the abnormal risk index of the dynamic object under the interaction effect are verified; The weighted evaluation formula calculates the abnormal risk index of the dynamic object in the target road by taking the abnormal risk index of the dynamic object under the interaction effect as a cooperative risk factor and the basic abnormal risk index of each dynamic object in the interaction causal matrix of the abnormal risk indexes of the dynamic objects in the target road.

[0015] Further, a traffic video abnormal control system based on a three-dimensional real scene map is proposed, which is used to implement the traffic video abnormal control method based on the three-dimensional real scene map, and comprises: The three-dimensional modeling module is used to acquire traffic monitoring video streams and three-dimensional modeling data of the target road based on traffic heterogeneous sensors, and establish a three-dimensional real scene modeling of the target road. ​​​​​The abnormality marking module is electrically connected with the three-dimensional modeling module, and is used for three-dimensional real scene modeling of the target road based on the traffic monitoring video stream of the target road, generating spatial mapping of the traffic monitoring video stream and the three-dimensional real scene modeling of the target road, and marking abnormal event information of a dynamic object in the target road. The abnormal risk quantification module is electrically connected with the abnormality marking module, and is used for real-time three-dimensional rendering according to the abnormal event information of the dynamic object in the target road, generating traffic state modeling of the three-dimensional real scene of the dynamic target road, and analyzing an abnormal risk index of the dynamic object in the target road. The control scheme generation module is electrically connected with the abnormal risk quantification module, and is used for visual marking of the three-dimensional real scene modeling of the target road based on the abnormal risk index of the dynamic object in the target road, determining an abnormal event type of the dynamic object, and generating a traffic abnormality control scheme of the target road.

[0016] Compared with the prior art, the beneficial effects of the present application are that: The present application provides a traffic video abnormality control scheme based on a three-dimensional real scene map, which integrates traffic heterogeneous sensor data to obtain traffic monitoring video stream and three-dimensional modeling data of a target road, and constructs a three-dimensional real scene model of the target road. In combination with spatial mapping of the traffic monitoring video stream and the three-dimensional modeling, abnormal target behavior data is marked in real time, and a dynamic traffic state model is generated through three-dimensional rendering to analyze risk indexes of abnormal targets. Finally, the abnormal event type is displayed through visual marking, and a targeted traffic abnormality scheme is generated. The beneficial effects of the present application are that: the traffic management efficiency is improved, the accident probability is reduced, and the traffic flow management and emergency response are optimized. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 It is a traffic video abnormality control method flowchart based on a three-dimensional real scene map; Figure 2 It is a traffic video abnormality control system framework diagram based on a three-dimensional real scene map. DETAILED DESCRIPTION

[0018] The following description is used to disclose the present application so that those skilled in the art can implement the present application. The preferred embodiments in the following description are only as examples, and other obvious modifications can be thought of by those skilled in the art.

[0019] REFERENCE Figure 1 As shown in the figure, a traffic video abnormality control method based on a three-dimensional real scene map comprises: S1, based on traffic heterogeneous sensors, obtaining traffic monitoring video stream and three-dimensional modeling data of a target road, and establishing three-dimensional real scene modeling of the target road; the three-dimensional modeling data comprises point cloud data, satellite images and street view data; The step S1 specifically comprises: Based on the multi-source data in the traffic monitoring three-dimensional modeling data of the target road, the ICP iterative closest point algorithm is used to eliminate the deviation for point cloud data registration, and the satellite image and the street view data are used as the texture source of the three-dimensional modeling to obtain an initialized target road three-dimensional real scene modeling data package; The initialized target road three-dimensional real scene modeling data package is subjected to data preprocessing; As further contents, the ICP algorithm needs to ensure sufficient initial matching accuracy when the point cloud data is registered, otherwise the error accumulation may be caused; when the satellite image and the street view data are used as the texture source, the accurate mapping of the texture and the three-dimensional model needs to be ensured, and when the GDAL is used for coordinate conversion, the coordinate system and the projection of different data sources need to be consistent to prevent the deviation in the conversion process; for the Poisson surface reconstruction algorithm, the normal vector estimation and sufficient point cloud density of the point cloud data are required to ensure the smoothness of the generated target road three-dimensional real scene modeling; The GDAL data conversion open source library is used to unify the coordinates of the traffic monitoring video stream of the target road and the initialized target road three-dimensional real scene modeling data package to obtain the multi-source fusion target road three-dimensional real scene modeling data; Based on the multi-source fusion target road three-dimensional real scene modeling data, the Poisson surface reconstruction algorithm is used to solve the gradient field of the implicit function and the normal vector field direction of the input point cloud data to be consistent to generate the target road three-dimensional real scene modeling; The ICP iterative closest point algorithm specifically comprises: In the formula, is a transformation matrix, is a point obtained by transforming the i-th source point of the source point, is the m-th point in the target point cloud, and m is the total amount of the point cloud data; The Poisson surface reconstruction algorithm specifically comprises:

[0020] In the formula, is a gradient operator, is the gradient of the implicit function at the point is the normal vector of the point at the point. In use, the above steps are combined with the contents in the steps:

[0021] ​​​As a further content, the fine registration is performed by the ICP algorithm, the splicing error of the multi-source point cloud data is significantly reduced, a high-precision data basis is provided for subsequent reconstruction, the Poisson reconstruction algorithm can reconstruct a smooth and detailed surface from the point cloud with a normal vector, can infer the partially occluded geometry, the coordinate unification is performed by using the GD library, the barriers between the data of different sensors are broken, the three-dimensional geometry (point cloud), high-altitude overhead information (satellite image), side view details (street view) and real-time dynamic flow (video) can be accurately superimposed on the same space reference, a unique and reliable data base is provided for subsequent space mapping and anomaly analysis, and the satellite image and the street view data are used as texture sources instead of using a single color or artificial mapping, so that the generated three-dimensional model maximally restores the real material, color and appearance of the target road. This not only improves the visual immersion, but more importantly, provides important real texture information for the AI algorithm to identify road markings, traffic signs and vehicle models; instead of a rough model, the step produces a high-precision, high-realism and multi-source fusion three-dimensional real scene modeling. This high-quality model is the absolute prerequisite for accurate mapping of video streams and three-dimensional space, and determines the accuracy and reliability of subsequent anomaly event marking and risk analysis.

[0022] S2, based on the traffic monitoring video stream of the target road and the three-dimensional real scene modeling of the target road, generating a space mapping of the traffic monitoring video stream and the three-dimensional real scene modeling of the target road, and marking the abnormal event information of the dynamic objects in the target road; Step S2 specifically comprises: Based on the traffic monitoring video stream of the target road, according to the three-dimensional real scene modeling of the target road, a plurality of two-dimensional and three-dimensional feature image data points are randomly selected from the traffic monitoring video stream of the target road, and are substituted into the cv2.solvePnP function to calculate and solve the camera extrinsic parameter, the world coordinate and the intrinsic parameter of the traffic monitoring video stream of the target road for space mapping, so as to obtain the spatial position information of the traffic monitoring video stream of the target road. According to the frame rate of the video stream, the traffic monitoring video stream of the target road is divided to obtain a traffic monitoring image frame sequence of the target road.

[0023] Step S2 further comprises: The spatial position information of the traffic monitoring video stream of the target road is used as the image attribute in the traffic monitoring image frame sequence of the target road, the Yolov5 algorithm is used to collect the dynamic object motion trajectory in the traffic monitoring image frame sequence of the target road, and the running trajectory feature data of each dynamic object is extracted; the running trajectory feature data includes: speed, acceleration, direction change; By using DeepSORT multi-target tracking, the trajectories of dynamic objects in adjacent frames of the trajectory feature data of each dynamic object are associated, and the position coordinates of dynamic objects in the complete data stream are marked to generate a sequence of motion object trajectory frames in the traffic monitoring image of the target road.

[0024] Step S2 also includes: Based on the Lstm long short-term memory network, an abnormal target behavior trajectory analysis model is established. The target road traffic monitoring image moving object trajectory frame sequence is used as input. According to several trajectory step lengths of the target road traffic monitoring image moving object trajectory history, the future moving object trajectory is predicted, and abnormal target behavior label data in the target road is generated. Based on the intrinsic and extrinsic parameters of traffic monitoring sensors on the target road, the two-dimensional spatial coordinates of abnormal target behavior label data in the target road are mapped to the three-dimensional real-scene model of the target road to obtain abnormal event information of dynamic objects in the target road. Specifically, the target behavior trajectory anomaly analysis model is as follows:

[0025] In the formula, The first in the target road j The first dynamic object t Information on abnormal events per unit of time, The first in the target road The first dynamic object t Hidden state per unit time The first in the target road The first dynamic object t Trajectory feature data per unit time, The first in the target road Weight matrix of a dynamic object The first in the target road Bias terms for a dynamic object It is the sigmoid activation function; When using it, combine the above steps with the content in the following: As a further content, by establishing the spatial mapping of the traffic monitoring video and the three-dimensional real scene model, and then using the YOLOv5 and DeepSORT algorithms to accurately extract and track the motion trajectory of the dynamic target, and combining the LSTM network to analyze the trajectory time sequence characteristics to intelligently identify abnormal behaviors, finally mapping the abnormal events back to the three-dimensional space for positioning. The beneficial effects of the scheme are to realize the leap from two-dimensional plane perception to three-dimensional measurable space, not only to make the traffic monitoring have the ability to quantitatively analyze the target speed, distance and other parameters, but also to improve the intelligent level and predictive foresight of abnormal event detection, and significantly enhance the situation awareness and decision support ability through three-dimensional intuitive visualization, and promote the development of traffic management to automation, precision and high efficiency.

[0026] S3, real-time three-dimensional rendering according to the abnormal event information of the dynamic object in the target road, generating traffic state modeling of the dynamic target road three-dimensional real scene, analyzing the abnormal risk index of the dynamic object in the target road; According to the observation window established per unit time, taking the abnormal event information of the dynamic object in the target road as the observation attribute, extracting the abnormal behavior feature data of the dynamic object in the target road; Based on the abnormal behavior feature data of the dynamic object in the target road, an abnormal behavior risk quantization function is established to calculate the abnormal risk index of the dynamic object in the initial target road; The abnormal behavior risk quantization function is specifically:

[0027] In the formula, is the abnormal risk index of the jth dynamic object in the target road in the tth unit time, is the motion speed value of the jth dynamic object in the target road in the tth unit time, is the acceleration value of the jth dynamic object in the target road in the tth unit time, is the direction change value of the jth dynamic object in the target road in the tth unit time, , , the motion speed value mean, the acceleration value mean, and the direction change value mean are represented in turn, , , are motion speed weight coefficients, motion speed weight coefficients, and direction change weight coefficients in turn.

[0028] Step S3 further comprises: According to the abnormal risk indicators of the dynamic objects in the initial target road, the interaction effect between the abnormal risk indicators of each independent dynamic object in the initial target road is verified according to the GAM generalized additive model, and an interaction causal matrix of the abnormal risk indicators of the dynamic objects in the initial target road is established; According to covariance analysis, the basic abnormal risk indicators of each dynamic object in the interaction causal matrix of the abnormal risk indicators of the dynamic objects in the initial target road and the abnormal risk indicators of the dynamic objects under the interaction effect are verified. The weighted evaluation formula calculates the abnormal risk indicators of the dynamic objects in the target road by taking the abnormal risk indicators of the dynamic objects under the interaction effect as the cooperative risk factors and the basic abnormal risk indicators of each dynamic object in the interaction causal matrix of the abnormal risk indicators of the dynamic objects in the target road.

[0029] S4, based on the abnormal risk indicators of the dynamic objects in the target road, three-dimensional real scene modeling visualization marking of the target road is performed, the abnormal event type of the dynamic object is determined, and a traffic abnormal control scheme of the target road is generated.

[0030] It should be noted that according to the abnormal event type, the traffic abnormal scheme of the target road is generated, and the exemplary emergency treatment scheme is according to the nature of the abnormal event, and the corresponding emergency response measures are started, including on-site investigation, traffic diversion, resource allocation, etc. At the same time, dynamically adjust the control strategy, optimize the traffic signal in real time according to the event development, adjust the traffic direction, set the bypass channel, and timely release the warning and notice through the information platform, so as to ensure the rapid recovery of order and minimize the impact on traffic.

[0031] In use, the contents in the above steps are combined: As further content, the present scheme combines traffic monitoring video stream and three-dimensional real scene modeling of the target road, uses Zhang Zhengyou calibration method to obtain the internal and external parameters of the traffic monitoring sensor, extracts the motion trajectory features (such as speed, acceleration, direction change) of the dynamic object based on Yolov5 algorithm, and uses LSTM network to analyze the abnormal behavior of the dynamic object, and maps the abnormal behavior label in two-dimensional space to the three-dimensional real scene of the target road based on the sensor parameters, and detects and marks the abnormal event of the dynamic object in real time. The beneficial effects of the present scheme are: it can realize multi-target road monitoring, abnormal behavior prediction and accurate positioning, and improve traffic management and safety.

[0032] Referring to Figure 2 Based on the same inventive concept of the traffic video abnormal control method based on a three-dimensional real scene map, a traffic video abnormal control system based on a three-dimensional real scene map is proposed, which comprises: The three-dimensional modeling module is used for acquiring traffic monitoring video stream and three-dimensional modeling data of a target road based on a traffic heterogeneous sensor, and establishing three-dimensional real scene modeling of the target road; The abnormality marking module is electrically connected with the three-dimensional modeling module, and is used for generating spatial mapping of the traffic monitoring video stream and the three-dimensional real scene modeling of the target road based on the traffic monitoring video stream and the three-dimensional real scene modeling of the target road, and marking abnormal event information of a dynamic object in the target road; The abnormal risk quantification module is electrically connected with the abnormality marking module, and is used for generating traffic state modeling of a dynamic target road three-dimensional real scene by real-time three-dimensional rendering according to the abnormal event information of the dynamic object in the target road, and analyzing abnormal risk indexes of the dynamic object in the target road; The control scheme generation module is electrically connected with the abnormal risk quantification module, and is used for determining an abnormal event type of the dynamic object by visualizing labeling of the target road three-dimensional real scene modeling based on the abnormal risk indexes of the dynamic object in the target road, and generating a traffic abnormality control scheme of the target road.

[0033] The basic principle, main features and advantages of the present application are shown and described above. It should be understood by those skilled in the art that the present application is not limited by the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection claimed by the present application is defined by the appended claims and their equivalents.

Claims

1.A traffic video anomaly management method based on a three-dimensional real scene map, characterized in that, The method comprises the following steps: S1, based on traffic heterogeneous sensors, acquiring traffic monitoring video stream and three-dimensional modeling data of the target road, and establishing three-dimensional real scene modeling of the target road; The three-dimensional modeling data includes: point cloud data, satellite images, street view data; S2, based on the traffic monitoring video stream of the target road and the three-dimensional real scene modeling of the target road, generating the spatial mapping of the traffic monitoring video stream and the three-dimensional real scene modeling of the target road, and marking the abnormal event information of the dynamic objects in the target road; S3, according to the abnormal event information of the dynamic objects in the target road, real-time three-dimensional rendering is performed to generate the traffic state modeling of the dynamic target road three-dimensional real scene, and the abnormal risk index of the dynamic objects in the target road is analyzed; S4, based on the abnormal risk index of the dynamic objects in the target road, visualizing the three-dimensional real scene modeling of the target road, determining the abnormal event type of the dynamic objects, and generating the traffic abnormal control scheme of the target road. 2.The traffic video anomaly management method based on three-dimensional real scene map according to claim 1, wherein, Step S1 specifically comprises: Based on the multi-source data in the traffic monitoring three-dimensional modeling data of the target road, using the ICP iterative closest point algorithm, the point cloud data is registered to eliminate deviation, and the satellite images and street view data are used as the texture source of three-dimensional modeling, to obtain the initialization target road three-dimensional real scene modeling data package; Data preprocessing is performed on the initialization target road three-dimensional real scene modeling data package; Using the GDAL data conversion open source library, the coordinates of the traffic monitoring video stream of the target road and the initialization target road three-dimensional real scene modeling data package are unified, and the multi-source fusion target road three-dimensional real scene modeling data is obtained; Based on the multi-source fusion target road three-dimensional real scene modeling data, using the Poisson surface reconstruction algorithm, the gradient field of the implicit function is solved to be consistent with the normal vector field direction of the input point cloud data, and the target road three-dimensional real scene modeling is generated; The ICP iterative closest point algorithm specifically comprises: ； In the formula, The transformation matrix is... To transform the matrix The point obtained by transforming the i-th source point. For the target point cloud There are 1 point, where m is the total amount of point cloud data; The Poisson surface reconstruction algorithm specifically comprises: ； In the formula, For gradient operators, implicit function At point gradient at, For point The normal vector at that location. 3.The traffic video anomaly management method based on three-dimensional real scene map according to claim 2, characterized in that, Step S2 specifically comprises: Based on the traffic monitoring video stream of the target road, according to the three-dimensional real scene modeling of the target road, randomly selecting a plurality of groups of two-dimensional and three-dimensional feature image data points in the traffic monitoring video stream of the target road, and substituting them into the cv2.solvePnP function to calculate the camera external parameter, world coordinate and internal parameter of the traffic monitoring video stream of the target road for spatial mapping, so as to obtain the spatial position information of the traffic monitoring video stream of the target road; According to the frame rate of the video stream, the traffic monitoring video stream of the target road is divided to obtain the traffic monitoring image frame sequence of the target road. 4.The traffic video anomaly management method based on three-dimensional real scene map according to claim 3, characterized in that, Step S2 further comprises: The spatial position information of the traffic monitoring video stream of the target road is used as the image attribute in the traffic monitoring image frame sequence of the target road, the Yolov5 algorithm is used to collect the motion trajectory of the dynamic object in the traffic monitoring image frame sequence of the target road, and the running trajectory feature data of each dynamic object is extracted; the running trajectory feature data includes: speed, acceleration, direction change; The running track of each dynamic object is associated by using DeepSORT multi-target tracking, and the position coordinates of the dynamic object in the complete data stream are marked to generate a traffic monitoring image motion object track frame sequence of the target road. 5.The traffic video anomaly management method based on three-dimensional real scene map according to claim 4, characterized in that, The step S2 further includes: An abnormal behavior track analysis model is established based on an Lstm long short-term memory network, the traffic monitoring image motion object track frame sequence of the target road is taken as input, the future motion object track is predicted according to a plurality of track steps of the traffic monitoring image motion object track history of the target road, and abnormal target behavior label data of the target road is generated; Based on the internal and external parameters of the traffic monitoring sensor of the target road, the two-dimensional space coordinates of the abnormal target behavior label data of the target road are mapped into the three-dimensional real scene modeling of the target road, and abnormal event information of the dynamic object in the target road is obtained; The target behavior track abnormality analysis model specifically includes: ； In the formula, is the abnormal event information of the i-th dynamic object in the target road at the j-th unit time, j t is the hidden layer state of the i-th dynamic object in the target road at the j-th unit time, t is the trajectory feature data of the i-th dynamic object in the target road at the j-th unit time, t is the weight matrix of the i-th dynamic object in the target road, is the bias term of the i-th dynamic object in the target road, is a sigmoid activation function;​​​​​​​​ An observation window is established according to a unit time, the abnormal event information of the dynamic object in the target road is taken as an observation attribute, and abnormal behavior feature data of the dynamic object in the target road is extracted; Based on the abnormal behavior feature data of the dynamic object in the target road, an abnormal behavior risk quantization function is established, and an abnormal risk index of the dynamic object in the initial target road is calculated; The abnormal behavior risk quantization function specifically includes: ； In the formula, is an abnormal risk index of the jth dynamic object in the target road at the tth unit time, is a motion speed value of the jth dynamic object in the target road at the tth unit time, is an acceleration value of the jth dynamic object in the target road at the tth unit time, is a direction change value of the jth dynamic object in the target road at the tth unit time, , , the motion speed value mean, the acceleration value mean, and the direction change value mean are represented in sequence, , , the motion speed weight coefficient, the motion speed weight coefficient, and the direction change weight coefficient are represented in sequence. 6.The traffic video anomaly management method based on three-dimensional real scene map according to claim 5, characterized in that, The step S3 further includes: According to the abnormal risk index of the dynamic object in the initial target road, the interaction influence effect between the abnormal risk indexes of each independent dynamic object in the initial target road is verified according to a GAM general additive model, and an interaction causal matrix of the abnormal risk index of the dynamic object in the initial target road is established; According to covariance analysis, the basic abnormal risk index of each dynamic object in the interaction causal matrix of the abnormal risk index of the dynamic object in the initial target road and the abnormal risk index of the dynamic object under the interaction influence effect are verified; The weighted evaluation formula takes the abnormal risk index of the dynamic object under the interaction influence effect as a cooperative risk factor and the basic abnormal risk index of each dynamic object in the interaction causal matrix of the abnormal risk index of the dynamic object in the target road, and calculates the abnormal risk index of the dynamic object in the target road. 7.A traffic video anomaly management system based on a three-dimensional real scene map, characterized in that, The method comprises the following steps: The three-dimensional modeling module is used for acquiring the traffic monitoring video stream and the three-dimensional modeling data of the target road based on the traffic heterogeneous sensor, and establishing the three-dimensional real scene modeling of the target road; The abnormal marking module is electrically connected with the three-dimensional modeling module, and the abnormal marking module is used for generating the space mapping of the traffic monitoring video stream and the three-dimensional real scene modeling of the target road based on the traffic monitoring video stream of the target road and the three-dimensional real scene modeling of the target road, and marking the abnormal event information of the dynamic object in the target road. The abnormal risk quantification module is electrically connected with the abnormal marking module, and is configured to perform real-time three-dimensional rendering according to the abnormal event information of the dynamic object in the target road, generate traffic state modeling of the three-dimensional real scene of the dynamic target road, and analyze the abnormal risk index of the dynamic object in the target road. The management and control scheme generation module is electrically connected with the abnormal risk quantification module, and is configured to perform three-dimensional real scene modeling visual marking of the target road based on the abnormal risk index of the dynamic object in the target road, determine the abnormal event type of the dynamic object, and generate a traffic abnormality management and control scheme of the target road.

Images