Collision detection method, system, medium and program product based on 3D projection modeling

Abstract

A collision detection method, system, medium and program product based on 3D projection modeling, in which the following steps are used: receiving an approach monitoring video; establishing a collision enclosing cuboid model of a vehicle; obtaining video information of other objects in the parking lot, and establishing other collision enclosing cuboid models corresponding to other objects; projecting the collision enclosing cuboid model and other collision enclosing cuboid models onto the ground plane to obtain the vehicle rectangle and other rectangles; projecting the vehicle rectangle and other rectangles onto the length and width to obtain four projection results; when any projection result does not contain an overlap point, the vehicle is determined to be in a safe state; when only three projection results contain an overlap point, the vehicle control computer issues a preset warning message; when all four projection results contain an overlap point, the vehicle is determined to have an accident. This reduces the probability of collision or scratching of cars in small and complex parking lots.

Description

Technical Field

[0001] The present application belongs to the field of electronic digital data processing, and in particular, relates to a collision detection method, system, medium and program product based on 3D projection modeling. Background Art

[0002] As the number of motor vehicles in cities continues to grow, accidents such as vehicle scratches and collisions in parking lots are frequent. As parking lots are small and have many vehicles, collisions are very likely to occur during vehicle entry, exit, and parking. Especially in low-light environments such as underground parking lots, drivers' ability to perceive the surrounding environment is limited, making it easier for vehicles to scratch and collide with each other.

[0003] At present, the anti-collision measures commonly used in parking lots mainly include static protection measures such as installing anti-collision strips on both sides of the parking spaces and setting up anti-collision facilities around the pillars. Although these measures can reduce the damage caused by collisions to a certain extent, it is difficult to effectively prevent the occurrence of collision accidents. At the same time, although some vehicles are equipped with auxiliary systems such as reversing radar and 360-degree panoramic images, these systems mainly rely on the detection results of on-board sensors, and still have certain limitations in spatial perception and collision warning, making it difficult to provide drivers with comprehensive and accurate warning information. Summary of the invention

[0004] The present application provides a collision detection method, system, medium and program product based on 3D projection modeling, which are used to reduce the probability of collision or scratch of cars in small and complex parking lots.

[0005] In a first aspect, the present application provides a collision detection method based on 3D projection modeling, which receives an entry monitoring video captured by a parking lot entrance and exit monitoring device when a vehicle enters a parking lot;

[0006] Establish a collision enclosure cuboid model of the vehicle based on the on-site monitoring video;

[0007] Obtain video information of other objects in the parking lot, and establish other collision enclosure cuboid models corresponding to other objects one by one based on the video information;

[0008] The collision enclosing cuboid model and other collision enclosing cuboid models are projected onto the ground plane respectively to obtain the vehicle rectangle corresponding to the collision enclosing cuboid model and other rectangles corresponding to the other collision enclosing cuboid models;

[0009] Project the vehicle rectangle and other rectangles onto the length and width to obtain four projection results;

[0010] When any projection result does not contain a coincident point, the vehicle is determined to be in a safe state;

[0011] When only three projection results contain overlapping points, the vehicle computer controlling the vehicle issues a preset warning message;

[0012] When the four projection results all contain overlapping points, it is determined that the vehicle has an accident.

[0013] By adopting the above technical solution, the approach monitoring video shot by the parking lot entrance and exit monitoring equipment is received to establish a collision enclosure cuboid model, and the three-dimensional spatial characteristics of the vehicle are accurately obtained. At the same time, the video information of other objects in the parking lot is obtained and the corresponding collision enclosure cuboid model is established, realizing the three-dimensional modeling of all objects in the parking lot. The three-dimensional collision enclosure cuboid model is projected onto the ground plane to obtain a two-dimensional rectangle, which reduces the computational complexity of spatial collision detection. The two-dimensional rectangle is projected four times in the length and width directions, and the distribution of the coincidence points of the one-dimensional projection reflects the relative position relationship between the vehicle and other objects. Based on the number of coincidence points, the vehicle state is divided into a safe state, a warning state and an accident state, and a hierarchical collision risk assessment mechanism is established. When the three projection results contain coincidence points, a warning message is issued in time, so that the driver can take effective avoidance measures. This dimensionality reduction detection method from three-dimensional to two-dimensional and then to one-dimensional improves the computational efficiency while ensuring the detection accuracy, and realizes real-time monitoring and early warning of vehicle collision risks in the parking lot.

[0014] In combination with some embodiments of the first aspect, in some embodiments, establishing a collision enclosure cuboid model of the vehicle according to the approach monitoring video specifically includes:

[0015] Perform noise reduction processing on the on-site surveillance video to obtain a denoised surveillance video;

[0016] Extracting vehicle point cloud information of the vehicle in the denoised surveillance video, and generating vehicle point cloud data based on the vehicle point cloud information;

[0017] Insert the vehicle point cloud data into the preset triangulated network to obtain a triangulated network model, which includes a number of triangles;

[0018] Use the first-order and second-order methods to count the vertices of all triangles and get the vertex values;

[0019] Input the vertex value into the covariance matrix calculation function group to obtain the covariance matrix;

[0020] Normalize the eigenvalues ​​of the covariance matrix to obtain the basis;

[0021] A collision enclosing rectangular parallelepiped model of the vehicle is established according to the basis and preset model building rules. The preset building rules are to determine the extreme points along each axial direction of the basis, determine the size of the collision enclosing rectangular parallelepiped model according to the extreme points, and use the axial direction of the basis as the direction of the collision enclosing rectangular parallelepiped model.

[0022] By adopting the above technical scheme, the noise reduction processing of the approach monitoring video reduces the interference information in the video and improves the extraction quality of the vehicle point cloud information. The vehicle point cloud data is extracted and generated from the denoised video, reflecting the spatial distribution characteristics of the vehicle's external contour. The point cloud data is inserted into the preset triangulated network to form a triangulated network model, realizing the discretization representation of the vehicle surface. The first-order and second-order methods are used to perform statistical analysis on the triangle vertices, retaining the local geometric features of the vehicle surface. The covariance matrix is ​​calculated by the covariance matrix calculation function group, reflecting the distribution law of the point cloud data in three-dimensional space. The characteristic roots of the covariance matrix are normalized to obtain the basis, and the main direction of the vehicle in space is determined. According to the basis and the preset establishment rules, a collision enclosing cuboid model is constructed, so that the cuboid model can fit the vehicle entity closely. This modeling method based on point cloud processing and main direction analysis improves the compactness of the model while ensuring the modeling accuracy, and provides a reasonable geometric simplified model for subsequent collision detection.

[0023] In conjunction with some embodiments of the first aspect, in some embodiments, the covariance matrix calculation function group is: In the function, is the mean, is the total number of triangles, , and are the coordinates of the three vertices of the i-th triangle, , and is the column vector of the ith triangle, is the covariance matrix.

[0024] By adopting the above technical solutions, the covariance matrix calculation function group establishes a mapping relationship from the vertices of the triangulated network model to the covariance matrix. The overall offset of the vertex distribution is eliminated by calculating the mean U of the triangle vertices. The column vectors A_i, B_i, and C_i are calculated using the difference between the vertex and the mean, and the relative spatial distribution information of the vertices is retained. The covariance matrix T is calculated based on the product and column vectors to quantify the distribution correlation of the point cloud data in different directions. The parameter n in the function group represents the total number of triangles, which realizes the statistical analysis of all sampling points on the vehicle surface. The processing of the triangle vertex coordinates a_i, b_i, and c_i retains the local features of the vehicle surface. The final covariance matrix contains the main feature information of the vehicle shape, and the quantitative relationship between the point cloud data and the main direction of the vehicle is established through matrix operations. This mathematical method based on covariance analysis provides a statistical description of the spatial characteristics of the vehicle, so that the construction of the collision encirclement cuboid model has a reliable mathematical basis.

[0025] In combination with some embodiments of the first aspect, in some embodiments, the vehicle rectangle and other rectangles are projected onto the length and width to obtain four projection results, specifically including:

[0026] The long side and the wide side of the vehicle rectangle are respectively used as the first projection axis and the second projection axis;

[0027] The long side and wide side of the other rectangular projections are used as the third projection axis and the fourth projection axis respectively;

[0028] The vehicle rectangle and other rectangular projections are projected onto the first projection axis, the second projection axis, the third projection axis and the fourth projection axis respectively to obtain four projection results.

[0029] By adopting the above technical solution, the long side and wide side of the vehicle rectangle are used as the first projection axis and the second projection axis, and the projection direction corresponding to the geometric characteristics of the vehicle itself is established. The long side and wide side of other rectangles are used as the third projection axis and the fourth projection axis, and the relative direction relationship between different vehicles is considered. The four projection axes correspond to the main geometric directions of vehicles and other objects, respectively, reflecting the spatial distribution characteristics of objects in different directions. The rectangle on the two-dimensional plane is projected onto the one-dimensional axis, and the complex plane overlap detection is converted into a simple line segment coincidence judgment. Projection detection is performed in four different directions to avoid the problem of missed detection that may be caused by single-direction projection. This multi-directional projection detection method reduces the computational complexity and realizes a comprehensive assessment of the collision risk on the two-dimensional plane through the combined analysis of one-dimensional projections.

[0030] In combination with some embodiments of the first aspect, in some embodiments, the steps of projecting the vehicle rectangle and other rectangular projections onto the first projection axis, the second projection axis, the third projection axis, and the fourth projection axis respectively, the projection process specifically includes: obtaining the coordinates of the first endpoint of the straight line where the projection axis is located and the second endpoint coordinates ;

[0031] Calculate the equation of the line based on the coordinates of the first endpoint and the second endpoint In and ,in: For the point to be projected , calculate its projection point on the projection axis ,in:

[0032] By adopting the above technical solution, the coordinates of the two end points of the projection axis are obtained to establish the equation of the line, and the projection coordinates of the point to be projected on the axis are calculated using the principle of vector operation, thereby ensuring the geometric accuracy of the projection result. This projection calculation method based on analytic geometry avoids the cumulative errors caused by coordinate transformation and numerical approximation in traditional projection methods, and improves the accuracy of collision detection. Through a standardized projection calculation process, the system can quickly calculate the precise projection position of any point to the projection axis, reducing collision misjudgments caused by projection calculation errors. This precise projection calculation method makes the system more reliable when judging the spatial relationship between vehicles and other objects, and improves the accuracy and timeliness of collision warnings.

[0033] In combination with some embodiments of the first aspect, in some embodiments, when the four projection results all contain overlapping points, after determining that the vehicle has an accident, the method further includes:

[0034] Obtaining relative motion trajectory data at the moment of collision;

[0035] Calculate the collision angle and relative speed based on the relative motion trajectory data;

[0036] The collision type is classified into rear-end collision, side collision or frontal collision according to the collision angle;

[0037] Set different dynamic warning areas based on collision type.

[0038] By adopting the above technical solution, the collision types are divided into rear-end collision, side collision and frontal collision according to the collision angle, so that the system can accurately identify different types of collision scenes. This collision classification method based on kinematic characteristics can effectively distinguish the dangerous characteristics under different collision forms, and provide a scientific basis for the subsequent setting of early warning areas. By accurately identifying the collision type, the system can more accurately assess the severity and potential danger of the accident, so as to take more targeted early warning measures. This intelligent collision classification mechanism improves the system's adaptability to different types of collisions and enhances the pertinence and effectiveness of early warning.

[0039] In conjunction with some embodiments of the first aspect, in some embodiments, different dynamic warning areas are set based on the collision type, specifically including:

[0040] Calculate the distance from the collision point to the geometric center of the two vehicles, and use the distance as a benchmark to establish a multi-level warning area around the collision point;

[0041] When it is a rear-end collision, the warning area is extended along the direction of vehicle travel; when it is a side collision, the warning area is extended perpendicular to the direction of vehicle travel; when it is a head-on collision, symmetrical warning areas are set.

[0042] By adopting the above technical solution, a multi-level warning area is established according to the distance from the collision point to the geometric center of the vehicle, and different warning area extension strategies are adopted for different types of collisions. For rear-end collisions, extending the warning area along the direction of vehicle travel can effectively cover the dangerous areas in front and behind the vehicle; for side collisions, extending the warning area perpendicular to the direction of travel can fully consider the characteristics of side collisions; for head-on collisions, setting symmetrical warning areas can fully cover the dangerous areas facing each other. This method of dynamically adjusting the warning area according to the type of collision makes the shape and size of the warning area more consistent with the actual collision risk, avoiding false alarms and missed alarms caused by improper setting of the warning area. By reasonably setting the shape and range of the warning area, the accuracy and practicality of the system warning are improved, making the warning more in line with the needs of actual collision scenarios.

[0043] In a second aspect, an embodiment of the present application provides a collision detection system based on 3D projection modeling, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and one or more processors call the computer instructions to enable the system to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0044] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, which, when executed on a system, causes the system to execute the method described in the first aspect and any possible implementation of the first aspect.

[0045] In a fourth aspect, an embodiment of the present application provides a computer program product, characterized in that when the computer program product runs on a system, the system executes the method described in any possible implementation manner in the first aspect.

[0046] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0047] 1. The present application provides a collision detection method based on 3D projection modeling, which receives the approach monitoring video shot by the parking lot entrance and exit monitoring equipment to establish a collision enclosed rectangular parallelepiped model, and accurately obtains the three-dimensional spatial characteristics of the vehicle. At the same time, the video information of other objects in the parking lot is obtained and the corresponding collision enclosed rectangular parallelepiped model is established, realizing the three-dimensional modeling of all objects in the parking lot. The three-dimensional collision enclosed rectangular parallelepiped model is projected onto the ground plane to obtain a two-dimensional rectangle, which reduces the computational complexity of spatial collision detection. The two-dimensional rectangle is projected four times in the length and width directions, and the relative position relationship between the vehicle and other objects is reflected by the distribution of the coincidence points of the one-dimensional projection. Based on the number of coincidence points, the vehicle state is divided into a safe state, a warning state and an accident state, and a hierarchical collision risk assessment mechanism is established. When the three projection results contain coincidence points, a warning message is issued in time, so that the driver can take effective avoidance measures. This dimensionality reduction detection method from three-dimensional to two-dimensional to one-dimensional improves the computational efficiency while ensuring the detection accuracy, and realizes real-time monitoring and early warning of vehicle collision risks in the parking lot.

[0048] 2. The present application provides a collision detection method based on 3D projection modeling, which obtains the coordinates of the two end points of the projection axis to establish a straight line equation, and uses the principle of vector operation to calculate the projection coordinates of the point to be projected on the axis, thereby ensuring the geometric accuracy of the projection result. This projection calculation method based on analytic geometry avoids the cumulative errors caused by coordinate transformation and numerical approximation in traditional projection methods, and improves the accuracy of collision detection. Through a standardized projection calculation process, the system can quickly calculate the precise projection position of any point to the projection axis, reducing collision misjudgments caused by projection calculation errors. This precise projection calculation method makes the system more reliable when judging the spatial relationship between vehicles and other objects, and improves the accuracy and timeliness of collision warnings.

[0049] 3. The present application provides a collision detection method based on 3D projection modeling, which divides the collision types into rear-end collisions, side collisions and frontal collisions according to the collision angle, so that the system can accurately identify different types of collision scenarios. This collision classification method based on kinematic characteristics can effectively distinguish the dangerous characteristics under different collision forms, and provide a scientific basis for the setting of subsequent warning areas. By accurately identifying the collision type, the system can more accurately assess the severity and potential dangers of the accident, so as to take more targeted warning measures. This intelligent collision classification mechanism improves the system's adaptability to different types of collisions and enhances the pertinence and effectiveness of the warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a flow chart of a collision detection method based on 3D projection modeling in an embodiment of the present application.

[0051] Figure 2 It is a flow chart of a method for setting a dynamic warning area based on collision type in an embodiment of the present application.

[0052] Figure 3 It is a schematic diagram of the physical device structure of a collision detection system based on 3D projection modeling provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification and appended claims of the present application, the singular expressions "one", "a kind of", "", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more of the listed items.

[0054] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0055] The following uses an embodiment and combines Figure 1 , a collision detection method based on 3D projection modeling in an embodiment of the present application is described:

[0056] See also Figure 1 , which is a flow chart of a collision detection method based on 3D projection modeling in an embodiment of the present application.

[0057] S101. When a vehicle enters a parking lot, receiving an entry monitoring video shot by a parking lot entrance and exit monitoring device;

[0058] In this step, the system needs to receive and obtain the entry monitoring video data taken by the parking lot entrance and exit monitoring equipment in real time when the vehicle enters the parking lot. These monitoring video data contain the image information of the vehicle entering the parking lot, which is the basis for subsequent 3D reconstruction and collision detection. The monitoring equipment usually uses a high-definition camera, which can provide sufficiently clear and stable video images to ensure the quality of vehicle image acquisition.

[0059] In order to achieve real-time reception of surveillance video data, the system can be connected and integrated with the parking lot's surveillance system. Through network communication protocols such as RTSP, ONVIF, etc., the system can establish a data transmission channel with the surveillance camera and receive the video stream data collected by the camera in real time according to the set video encoding format and transmission parameters. At the same time, in order to ensure the stability and reliability of data transmission, the system can also adopt a buffering mechanism to maintain a video frame buffer at the receiving end for temporary storage and smoothing of video streams to avoid the impact of network jitter on image quality.

[0060] In actual applications, the monitoring equipment in parking lots may be of various types and models. In order to improve the compatibility and adaptability of the system, the system can also use video interface conversion and adaptation technology to support the access of video sources of different formats and protocols. For example, the analog video signal is converted into a digital video stream through a video decoder, and the private protocol is converted into a standard protocol through a protocol gateway. In addition, in view of the large-scale and high-concurrency characteristics of video data, the system can also introduce distributed computing frameworks such as Apache Storm and Spark Streaming to achieve efficient reception, storage and calculation of video data, and improve the performance and scalability of the system.

[0061] S102, establishing a collision enclosure cuboid model of the vehicle according to the approach monitoring video;

[0062] The system establishes a collision enclosure cuboid model of the vehicle based on the approaching surveillance video. Specifically: the approaching surveillance video is subjected to noise reduction processing to obtain a denoised surveillance video;

[0063] Extracting vehicle point cloud information of the vehicle in the denoised surveillance video, and generating vehicle point cloud data based on the vehicle point cloud information;

[0064] Insert the vehicle point cloud data into the preset triangulated network to obtain a triangulated network model, which includes a number of triangles;

[0065] Use the first-order and second-order methods to count the vertices of all triangles and get the vertex values;

[0066] Input the vertex value into the covariance matrix calculation function group to obtain the covariance matrix;

[0067] Normalize the eigenvalues ​​of the covariance matrix to obtain the basis;

[0068] A collision enclosing rectangular parallelepiped model of the vehicle is established according to the basis and preset model building rules. The preset building rules are to determine the extreme points along each axial direction of the basis, determine the size of the collision enclosing rectangular parallelepiped model according to the extreme points, and use the axial direction of the basis as the direction of the collision enclosing rectangular parallelepiped model.

[0069] Among them, the covariance matrix calculation function group is: In the function, is the mean, is the total number of triangles, , and are the coordinates of the three vertices of the i-th triangle, , and is the column vector of the ith triangle, is the covariance matrix.

[0070] In this step, the system needs to perform 3D modeling of the vehicle entering the parking lot based on the approach monitoring video data obtained in the previous step, and generate a collision enclosure cuboid model that can approximate the shape and size of the vehicle. This model is a simplified 3D geometric body. Using a cuboid to enclose the main contour of the vehicle can not only reduce the complexity of the model, but also better describe the collision characteristics of the vehicle, which is an important basis for subsequent collision detection.

[0071] In order to establish the collision enclosure cuboid model, the system first needs to pre-process the approach monitoring video and extract the three-dimensional point cloud information of the vehicle. Specifically, the system can use computer vision algorithms such as video noise reduction, background difference, and target detection to segment the vehicle area from the video frame sequence and obtain the two-dimensional contour and pixel point set of the vehicle. Then, stereo vision technology such as binocular vision and structured light is used to reconstruct the vehicle image in three dimensions to generate the vehicle's three-dimensional point cloud data, in which each point contains its three-dimensional spatial coordinate information.

[0072] After obtaining the vehicle point cloud data, the system needs to further reconstruct its surface and approximate its shape to obtain a compact and regular three-dimensional model. To this end, the system can use triangulated meshing algorithms, such as Delaunay triangulation, BallPivoting, etc., to embed the vehicle point cloud data into a preset triangular mesh to generate a triangulated mesh model. Then, the triangulated mesh model is subjected to vertex analysis and statistics, and its local geometric features are estimated by calculating the first-order and second-order derivatives of each vertex, and the covariance matrix of the vertex is constructed. Next, the covariance matrix is ​​subjected to eigenvalue decomposition to obtain its eigenvectors, and the eigenvectors are normalized to form a set of orthogonal bases. Finally, according to the basis vectors and the preset cuboid construction rules, such as taking the extreme points of the vertices on each basis vector as the vertices of the cuboid, the position, direction and size of the collision enclosing cuboid model are determined.

[0073] In practical applications, due to the diversity of vehicle models and appearances, establishing a unified and robust collision bounding box model faces many challenges. In order to further improve the adaptability and accuracy of the model, the system can also introduce machine learning methods to automatically learn and optimize the cuboid construction rules and adjust the model parameters online through training on a large number of vehicle model sample data. For example, statistical shape modeling methods such as active shape models and support vector shape description are used to characterize the intrinsic distribution law of vehicle shapes. In addition, the system can also use prior knowledge of the vehicle, such as the geometric features of significant parts such as wheels and windows, to assist in guiding the 3D reconstruction process and improve the semantic consistency and robustness of the model.

[0074] S103, obtaining video information of other objects in the parking lot, and establishing other collision enclosing cuboid models corresponding to other objects one by one according to the video information;

[0075] In this step, the system needs to acquire and analyze the video information of other objects in the parking lot and establish corresponding collision enclosure cuboid models for other objects. Other objects mainly refer to fixed facilities and temporary obstacles in the parking lot, such as parking spaces, pillars, gates, other parked vehicles, etc. There is also a potential risk of collision between these objects and vehicles, and they need to be included in the scope of collision detection to improve the comprehensiveness and reliability of detection.

[0076] Similar to the establishment of the vehicle model, the system first needs to obtain the video surveillance data of other objects in the parking lot, connect and integrate with the data interface of the monitoring system, and receive and retrieve the video information of other objects in real time. Since the spatial distribution of other objects in the parking lot is relatively scattered, it may require the coordinated observation of multiple surveillance cameras. Therefore, the system needs to synchronize and align multi-source video data to establish a unified space-time coordinate system. At the same time, considering the diversity of attributes such as shape, size, texture, etc. of different objects, the system needs to adopt adaptive strategies for different types of objects when modeling to improve the accuracy and efficiency of modeling.

[0077] After obtaining video information of other objects, the system will use video analysis and 3D reconstruction technology similar to vehicle modeling to extract the 3D point cloud data of the objects, construct a triangulated network model, and generate a corresponding collision enclosure cuboid model. Since other objects are usually simpler and more regular in shape than vehicles, some simplified strategies can be used in the modeling process, such as using cylindrical enclosures for columns and cuboid enclosures for gates, to improve computational efficiency while ensuring modeling accuracy. In addition, for some stationary fixed facilities, such as parking spaces, 3D laser scanning and other technologies can be used to directly obtain their high-precision point cloud models, eliminating the process of video reconstruction.

[0078] S104, projecting the collision enclosing cuboid model and other collision enclosing cuboid models onto the ground plane respectively to obtain the vehicle rectangle and other rectangles;

[0079] In this step, the system needs to project the vehicle collision enclosure cuboid model and other object collision enclosure cuboid model established in the above step onto the ground to obtain the corresponding two-dimensional plane rectangle. This step is to prepare for the subsequent collision detection. By projecting the collision enclosure in three-dimensional space onto a two-dimensional plane, the computational complexity of collision detection can be simplified and the detection efficiency can be improved. At the same time, since the ground is the common motion plane for vehicles and other objects, collision analysis on the ground plane can better reflect the relative position relationship and collision risk between them.

[0080] Specifically, the system first needs to determine a unified ground reference plane, and the ground of the parking lot can usually be selected as the reference plane. Then, for each collision-enclosing rectangular model, the system needs to calculate the projection coordinates of its vertices on the ground plane. This process can be achieved through three-dimensional space geometric transformation, such as using perspective projection or orthogonal projection to transform the vertices of the rectangular block from the three-dimensional space coordinate system to the two-dimensional ground coordinate system. During the projection process, the orientation and tilt angle of the rectangular block need to be considered to ensure that the projected figure can accurately reflect its coverage area on the ground. The specific formula and parameters of the projection transformation can be calculated and optimized based on information such as the internal and external parameters of the camera and the normal vector of the ground plane.

[0081] S105, projecting the vehicle rectangle and other rectangles onto the length and width to obtain four projection results;

[0082] The system projects the vehicle rectangle and other rectangles onto the length and width to obtain four projection results. Specifically, the long side and the wide side of the vehicle rectangle are used as the first projection axis and the second projection axis respectively;

[0083] The long side and wide side of the other rectangular projections are used as the third projection axis and the fourth projection axis respectively;

[0084] The vehicle rectangle and other rectangular projections are projected onto the first projection axis, the second projection axis, the third projection axis and the fourth projection axis respectively to obtain four projection results.

[0085] The projection process specifically includes: obtaining the coordinates of the first endpoint of the line where the projection axis is located and the second endpoint coordinates ;

[0086] Calculate the equation of the line based on the coordinates of the first endpoint and the second endpoint In and ,in: For the point to be projected , calculate its projection point on the projection axis ,in:

[0087] In this step, the system needs to further project and decompose the vehicle rectangle and other object rectangles obtained in the previous step to obtain their one-dimensional projection results in the length and width directions. The purpose of this step is to further simplify the collision detection problem on the two-dimensional plane into an overlap judgment problem on a one-dimensional line segment. By comparing the projection intervals of the rectangles in different directions, it is possible to quickly determine whether there is a possibility of collision between them. At the same time, through the results of the four projections, the collision situation can be cross-validated from different directions to improve the reliability of detection.

[0088] Specifically, the system first needs to determine four projection axes, namely the long side of the vehicle rectangle, the wide side of the vehicle rectangle, the long side of the other object rectangle, and the wide side of the other object rectangle. These four projection axes represent the main directions of the vehicle and other objects on the plane. By analyzing the projection of the rectangle in these directions, the positional relationship between them can be fully characterized. The projection axis can be obtained through vector operations based on the vertex coordinates of the rectangle, such as using the line connecting the midpoints of two opposite sides as the direction vector of the projection axis.

[0089] After determining the projection axis, the system needs to project the vehicle rectangle and other object rectangles onto each projection axis to obtain the corresponding one-dimensional projection interval. The projection process can be implemented through analytical geometry methods, such as vector dot product, line segment intersection, etc. The specific steps are as follows: First, according to the coordinates of the endpoints of the projection axis, establish the parametric equation of the projection axis line; then, for each vertex of the rectangle, calculate the coordinates of its projection point on the projection axis; finally, find the maximum and minimum values ​​of the coordinates of all projection points to obtain the projection interval of the rectangle on the projection axis. Repeat the above steps to obtain the projection results of the vehicle rectangle and other object rectangles on the four projection axes.

[0090] It should be noted that in the actual projection process, since the rectangles of vehicles and other objects may be rotated and tilted, the calculation of the endpoints of the projection interval may involve complex operations such as coordinate rotation transformation and trigonometric functions. In order to simplify the amount of calculation and improve the efficiency of projection, the system can try to align the long and wide sides of the collision enclosing cuboid model with the axis of the scene coordinate system when establishing it, reducing the complexity of the rotation transformation. For some difficult alignment situations, the system can also use numerical approximation, table lookup and other methods to reduce the amount of calculation while ensuring accuracy.

[0091] S106, when any projection result does not contain a coincidence point, it is determined that the vehicle is in a safe state;

[0092] In this step, the system needs to determine whether the vehicle is in a safe state based on the four projection results obtained in the previous step. If there are no overlapping points between the projection intervals of the vehicle rectangle and other object rectangles on any projection axis, that is, the two intervals are completely separated without intersection, then it can be considered that there is no risk of collision between the vehicle and other objects, and the vehicle is in a safe state. This judgment rule conforms to the basic principle of spatial geometry, that is, if two objects do not overlap in space, then they will not overlap in any projection direction either.

[0093] In specific implementation, the system can quickly determine whether there is an overlap between the two projection intervals through interval comparison operations. Taking the projection interval of the vehicle rectangle as [a1, a2] and the projection interval of other object rectangles as [b1, b2] as an example, when a2 < b1 or b2 < a1, the two intervals have no overlap, and there is no possibility of collision between the vehicle and other objects on this projection axis. The system sequentially compares the intervals of the four projection axes. If there is no interval overlap on all projection axes, it is considered that the vehicle as a whole is in a safe state and can drive and park normally.

[0094] S107. When there are only three projection results containing overlapping points, control the vehicle's in-vehicle computer to issue a preset warning message;

[0095] In this step, the system needs to determine whether the vehicle is in a collision warning state based on the occurrence of overlapping points in the four projection results. If there are overlapping intervals on only three projection axes, that is, the projections of the vehicle rectangle and other object rectangles have intersections in these three directions, and are completely separated in the remaining one direction, it is considered that there is a certain possibility of collision between the vehicle and other objects, but it has not reached the level of actual collision. In this case, the system will send a collision warning message to the driver through the in-vehicle terminal, prompting the driver to take operations such as decelerating and steering to avoid colliding with the object in front.

[0096] The specific way to issue the collision warning message can be adapted according to the type and configuration of the in-vehicle terminal. For example, the system can pop up a text prompt box on the in-vehicle central control screen, displaying warning content such as "Obstacle ahead, please slow down"; it can also play a voice prompt message through the in-vehicle audio, such as "There is an obstacle 1 meter ahead, please pay attention to safety"; it can also present the position and distance information of the obstacle in the form of icons, symbols, etc. through the indicator lights on the instrument panel or the head-up display. At the same time, the system can also dynamically adjust the presentation method and intensity of the warning message according to the relative distance and speed between the vehicle and the obstacle. For example, the closer the distance and the faster the speed, the larger the font of the warning message will be, the redder the color will be, and the voice will become more urgent, etc., to attract the driver's attention.

[0097] S108. When the four projection results all contain overlapping points, it is determined that the vehicle has an accident.

[0098] In this step, the system needs to determine whether the vehicle has collided based on the four projection results obtained above. Since in step S105, the system has projected the collision bounding box models of the vehicle and other objects onto the four axes where the long sides and wide sides of the two objects are located, by analyzing the overlap of the four projections, it is possible to accurately infer whether a collision has occurred between the two objects.

[0099] Specifically, if the four projection results all contain overlapping points, that is, the vehicle and other objects have overlapping parts on the four projection axes, then it can be determined that the two objects must have contacted or collided in three-dimensional space. This is because only when two three-dimensional objects are truly in contact will they overlap on the projection surface at any angle. Conversely, if the projections of the two objects do not intersect on a certain projection surface, then it can be confirmed that they are separated in space and have not collided.

[0100] Therefore, the reliability of collision detection can be significantly improved by combining the four projection results for collision inference. The system can design a logic judgment matrix to combine and analyze the four projection results and give a confidence score for the occurrence of a collision based on different overlap situations. When the confidence exceeds a preset threshold, an accident alarm can be triggered to notify management personnel to handle the situation on site.

[0101] In the above embodiment, the approach monitoring video shot by the parking lot entrance and exit monitoring equipment is received to establish a collision enclosed rectangular parallelepiped model, and the three-dimensional spatial characteristics of the vehicle are accurately obtained. At the same time, the video information of other objects in the parking lot is obtained and the corresponding collision enclosed rectangular parallelepiped model is established, so as to realize the three-dimensional modeling of all objects in the parking lot. The three-dimensional collision enclosed rectangular parallelepiped model is projected onto the ground plane to obtain a two-dimensional rectangle, which reduces the computational complexity of spatial collision detection. The two-dimensional rectangle is projected four times in the length and width directions, and the distribution of the coincidence points of the one-dimensional projection reflects the relative position relationship between the vehicle and other objects. Based on the number of coincidence points, the vehicle state is divided into a safe state, a warning state and an accident state, and a hierarchical collision risk assessment mechanism is established. When the three projection results contain coincidence points, a warning message is issued in time, so that the driver can take effective avoidance measures. This dimensionality reduction detection method from three-dimensional to two-dimensional and then to one-dimensional improves the computational efficiency while ensuring the detection accuracy, and realizes the real-time monitoring and warning of the collision risk of vehicles in the parking lot.

[0102] After determining that a vehicle has an accident, in order to further evaluate the collision condition and take corresponding safety protection measures, the embodiment of the present application also provides a method for setting a dynamic warning area based on the collision type. This method analyzes the motion characteristics at the moment of collision and adopts differentiated warning strategies for different types of collisions, thereby improving the safety protection effect after the collision. Figure 2 , a method for setting a dynamic warning area based on collision type in an embodiment of the present application is described:

[0103] See also Figure 2 , which is a flow chart of a method for setting a dynamic warning area based on collision type in an embodiment of the present application.

[0104] S201, obtaining relative motion trajectory data at the time of collision, and calculating the collision angle and relative speed based on the relative motion trajectory data;

[0105] In this step, the system needs to obtain the vehicle's motion state data when the collision occurs, as a basis for subsequent analysis of the collision type and setting of warning areas. Specifically, the system collects the position, speed, acceleration and other dynamic parameters of the collision vehicle in real time through on-board sensors and vehicle network communications, and uploads these parameter data to the cloud platform for aggregation and processing.

[0106] In order to accurately describe the collision process, the system needs to focus on the relative motion trajectory between the vehicles. By comparing and analyzing the motion parameters of the two vehicles, their relative position changes and relative speed changes at the moment of collision can be obtained, thereby inferring the angle and intensity of the collision. The calculation of the relative motion trajectory can be based on classic kinematic models, such as uniform linear motion, uniformly accelerated linear motion, etc., or higher-order kinematic equations can be used, such as considering the influence of factors such as air resistance and road friction.

[0107] When acquiring relative motion trajectory data, the system also needs to pay attention to the real-time and continuity of the data. The collision process usually occurs in a very short period of time, so the frequency of data sampling needs to be high enough to ensure that the key moment of the collision is captured. At the same time, since the collision will cause impact and interference to the on-board sensors and communication systems, data may be lost or distorted. In order to cope with this situation, the system can adopt a multi-source heterogeneous data fusion method, comprehensively utilize various types of sensor data such as vehicle-mounted, road test, video, etc., and ensure the integrity and reliability of the trajectory data through cross-validation and compensation correction. In addition, the system can also introduce data preprocessing technologies such as trajectory smoothing and outlier detection to further improve the quality and accuracy of trajectory data.

[0108] Based on the relative motion trajectory data, the system needs to focus on calculating two key parameters: collision angle and relative speed. The collision angle indicates the relative motion direction of the two vehicles at the time of collision, which determines the type of collision and the distribution of impact force. The system can obtain the collision angle by calculating the angle between the velocity vectors of the two vehicles at the collision point. The relative speed indicates the relative motion rate of the two vehicles at the moment of collision, which directly determines the intensity and degree of damage of the collision. The relative speed can be calculated by the modulus value of the speed difference between the two vehicles at the collision point.

[0109] S202, classifying the collision type into a rear-end collision, a side collision, or a frontal collision according to the collision angle;

[0110] In this step, the system needs to classify the collision accident according to the collision angle calculated in the previous step. The collision type is the primary basis for formulating safety protection strategies. Different types of collisions have significant differences in damage mechanism, degree of danger, etc., and targeted early warning and disposal measures need to be taken. This embodiment focuses on three common types of collisions: rear-end collision, side collision and head-on collision.

[0111] Specifically, the system first needs to set the judgment thresholds for different collision types based on the range of collision angles. Generally speaking, when the collision angle is close to 0 degrees or 180 degrees, that is, the two vehicles are moving in the same or opposite directions, it can be judged as a rear-end collision. When the collision angle is close to 90 degrees, that is, the two vehicles are in a vertical motion state, it can be judged as a side collision. When the collision angle is close to 180 degrees, that is, the two vehicles are in a head-on motion state, it can be judged as a head-on collision. The setting of the threshold needs to comprehensively consider the complexity of actual road traffic and the strict requirements for safety protection. It is necessary to divide the collision types as accurately as possible, and avoid missing certain special working conditions due to an overly narrow threshold range.

[0112] S203, calculating the distance from the collision point to the geometric center of the two vehicles, and establishing a multi-level warning area around the collision point based on the distance;

[0113] In this step, the system needs to set up a reasonable warning area around the accident vehicle based on the location of the collision. The purpose of the warning area is to promptly alert and evacuate surrounding vehicles and pedestrians after the accident to prevent secondary accidents. A reasonable warning area needs to fully consider factors such as the location of the collision and the inertia of the vehicle. It should not only cover the main areas affected by the accident, but also avoid excessive diffusion and affect normal traffic.

[0114] The system first needs to calculate the distance from the collision point to the geometric center of the two vehicles. The geometric center of the vehicle is usually located on the axis of symmetry of the vehicle body and is the center of mass of the vehicle. The distance between the collision point and the geometric center can reflect the degree of eccentricity of the collision. The greater the distance, the more the collision point deviates from the center of the vehicle, and the more likely the vehicle will roll over, rotate, or lose control. The system can directly calculate this distance value through the three-dimensional model of the vehicle and the coordinates of the collision point.

[0115] According to the distance, the system will establish multiple levels of warning areas around the collision point. The warning area can be in the form of concentric circles or ellipses, expanding outward from the collision point. Different levels of warning areas correspond to different levels of danger and protection requirements. For example, the innermost warning area represents the high-risk area for accidents, which needs to be blocked immediately and prohibits external vehicles and pedestrians from entering; the middle-layer warning area represents the accident impact area, which needs to guide vehicles to detour and give way to rescue vehicles; the outermost warning area represents the accident warning area, and traffic signs and broadcasts are used to remind passing vehicles to slow down and avoid.

[0116] When setting the warning area, the system can use an adaptive area division algorithm to dynamically adjust the size and shape of the warning area according to factors such as the severity of the collision and the type and size of the vehicle. For example, for large buses or trucks, the collision inertia is large, so a larger warning area needs to be set; while for small passenger cars, the impact of the accident is relatively small, and the warning area can be appropriately reduced. In addition, the system can also consider the actual layout of the road and use the isolation belt, green belt and other facilities of the road to physically isolate the warning area to avoid excessive spread of warning information.

[0117] S204. When it is a rear-end collision, extend the warning area along the direction of vehicle travel; when it is a side collision, extend the warning area perpendicular to the direction of vehicle travel; when it is a head-on collision, set symmetrical warning areas.

[0118] In this step, the system needs to optimize the shape and direction of the warning area according to the previously identified collision type. Different types of collisions have obvious differences in vehicle forces, movement trends, etc., and accordingly, their impact areas on vehicle and personnel safety are also different. By extending and adjusting the warning area in a typological manner, collision protection can be implemented more accurately and effectively, reducing the secondary risk of accidents.

[0119] Specifically, when a rear-end collision is detected, the system will appropriately extend the longitudinal length of the warning area along the direction of the vehicle's travel. This is because in a rear-end collision, the rear vehicle is subjected to a large longitudinal impact force, and its inertia may push the vehicle to continue moving forward for a distance, posing a continuous threat to the vehicle and pedestrians in front. Therefore, it is necessary to reserve a sufficient buffer area in the direction of the vehicle's advance to guide the vehicle in front to slow down or change lanes in advance to avoid a secondary rear-end collision.

[0120] When a side collision is detected, the system will expand the lateral width of the warning area in a direction perpendicular to the direction of the vehicle's travel. Unlike rear-end collisions, side collisions usually occur at intersections or lane change areas where vehicles cross each other, and the vehicle will be subjected to a strong lateral impact force. This may cause the vehicle to slip, rotate, or other uncontrolled states, and its movement trajectory has a large degree of uncertainty. Therefore, the warning area for side collisions needs to cover a wider area on both sides of the vehicle to remind vehicles in adjacent lanes on the left and right to maintain a safe distance and pass carefully.

[0121] When a head-on collision is detected, the system usually sets a symmetrical warning area around the collision point. A head-on collision is the most serious type of collision. Two vehicles collide head-on at a high relative speed, resulting in huge energy release and body deformation. After the collision, the two vehicles may rebound or move sideways in opposite directions due to inertia, and their movement direction and distance are difficult to accurately estimate. Therefore, the warning area for a head-on collision usually has a centrally symmetrical or axisymmetric shape, and sufficient safety margins need to be reserved in the longitudinal and lateral directions of the vehicle.

[0122] In the above embodiment, the relative motion trajectory data at the moment of collision is analyzed, the collision angle and relative speed are calculated, and a collision type classification mechanism is established. The collision types are divided into rear-end collision, side collision and frontal collision according to the collision angle, so that the system can accurately identify different types of collision scenes. This collision classification method based on kinematic characteristics can effectively distinguish the dangerous characteristics under different collision forms, and provide a scientific basis for the subsequent setting of early warning areas. By accurately identifying the collision type, the system can more accurately assess the severity and potential dangers of the accident, so as to take more targeted early warning measures. This intelligent collision classification mechanism improves the system's adaptability to different types of collisions and enhances the pertinence and effectiveness of the early warning.

[0123] The following describes the system in the embodiment of the present invention from the perspective of hardware processing. Figure 3 , which is a schematic diagram of the physical device structure of a collision detection system based on 3D projection modeling provided in an embodiment of the present application.

[0124] It should be noted that Figure 3The structure of the system shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0125] like Figure 3 As shown, the system includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage part 308 to the random access memory (RAM) 303, such as executing the method in the above embodiment. In RAM 303, various programs and data required for system operation are also stored. CPU 301, ROM 302 and RAM 303 are connected to each other through a bus 304. Input / output (I / O) interface 305 is also connected to bus 304.

[0126] The following components are connected to the I / O interface 305: an input section 306 including a camera, an infrared sensor, etc.; an output section 307 including a liquid crystal display (LCD) and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed so that a computer program read therefrom is installed into the storage section 308 as needed.

[0127] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 309, and / or installed from a removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, various functions defined in the present invention are performed.

[0128] It should be noted that the computer-readable medium shown in the embodiment of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable computer program. Such a propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing.

[0129] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Among them, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0130] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the system described in the above embodiment; or may exist independently without being assembled into the system. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of a system, the system implements the method provided in the above embodiment.

[0131] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

[0132] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0133] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on the computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server, a data center, etc. that contains one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk), etc.

[0134] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. A collision detection method based on 3D projection modeling, characterized in that: include: When a vehicle enters a parking lot, receiving the entry monitoring video captured by the parking lot entrance and exit monitoring equipment; Establishing a collision enclosure cuboid model of the vehicle according to the approach monitoring video specifically includes: Performing noise reduction processing on the entry surveillance video to obtain a denoised surveillance video; Extracting vehicle point cloud information of the vehicle in the denoised surveillance video, and generating vehicle point cloud data based on the vehicle point cloud information; Inserting the vehicle point cloud data into a preset triangulated network to obtain a triangulated network model, wherein the triangulated network model includes a plurality of triangles; Using first-order and second-order methods to count the vertices of all the triangles to obtain vertex values; inputting the vertex values ​​into a covariance matrix calculation function group to obtain a covariance matrix; Normalizing the characteristic roots of the covariance matrix to obtain a basis; Establishing a collision enclosing cuboid model of the vehicle according to the basis and a preset model building rule, wherein the preset model building rule is to determine a polar vertex along each axial direction of the basis, determine a size of the collision enclosing cuboid model according to the polar vertex, and use the axial direction of the basis as the direction of the collision enclosing cuboid model; Acquire video information of other objects in the parking lot, and establish other collision enclosing cuboid models corresponding one-to-one to the other objects according to the video information; Respectively projecting the collision enclosing cuboid model of the vehicle and the other collision enclosing cuboid models onto a ground plane to obtain a vehicle rectangle corresponding to the collision enclosing cuboid model of the vehicle and other rectangles corresponding to the other collision enclosing cuboid models; Project the vehicle rectangle and the other rectangles onto the length and width to obtain four projection results; When any of the projection results does not contain a coincidence point, it is determined that the vehicle is in a safe state; When there are only three projection results containing the coincidence point, the vehicle computer of the vehicle is controlled to issue a preset warning message; When the four projection results all contain the coincidence point, it is determined that the vehicle has an accident.

2. The method according to claim 1, characterized in that: The vehicle rectangle and the other rectangles Projecting onto the length and width, we get four projection results, including: The long side and the wide side of the vehicle rectangle are used as the first projection axis and the second projection axis respectively; The long side and the wide side of the other rectangular projection are used as the third projection axis and the fourth projection axis respectively; The vehicle rectangle and the other rectangles are projected onto the first projection axis, the second projection axis, the third projection axis and the fourth projection axis respectively to obtain four projection results.

3. The method according to claim 2, characterized in that The step of projecting the vehicle rectangle and the other rectangles onto the first projection axis, the second projection axis, the third projection axis and the fourth projection axis respectively, the projection process specifically includes: obtaining the first endpoint coordinates T1 (x1, y1) and the second endpoint coordinates T2 (x2, y2) of the straight line where the projection axis is located; The k and the b in the straight line equation y=kx+b are calculated according to the first endpoint coordinates and the second endpoint coordinates, wherein: For the point to be projected T3(x3, y3), calculate its projection point T0(x0, y0) on the projection axis, where:

4. The method according to claim 1, characterized in that: When the four projection results all contain the coincidence point, it is determined that the vehicle has an accident, and the method further includes: Obtaining relative motion trajectory data at the moment of collision; Calculating the collision angle and relative speed based on the relative motion trajectory data; The collision type is divided into a rear-end collision, a side collision or a frontal collision according to the collision angle; and different dynamic warning areas are set based on the collision type.

5. The method according to claim 4, characterized in that The setting of different dynamic warning areas based on the collision type specifically includes: Calculating the distance between the collision point and the geometric center of the two vehicles, and establishing a multi-level warning area around the collision point based on the distance; When it is a rear-end collision, the warning area is extended along the vehicle's travel direction; when it is a side collision, the warning area is extended perpendicular to the vehicle's travel direction; when it is a head-on collision, symmetrical warning areas are set.

6. A collision detection system based on 3D projection modeling, characterized in that: The system comprises: One or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the system to execute the method as described in any one of claims 1-5.

7. A computer-readable storage medium comprising instructions, characterized in that: When the instructions are executed on a system, the system is caused to execute the method according to any one of claims 1 to 5.

8. A computer program product, characterized in that When the computer program product is run on a system, the system is caused to execute the method according to any one of claims 1 to 5.

Images