A dynamic collision detection method and device based on geometric model topological relation

Abstract

The application discloses a kind of dynamic collision detection method and device based on geometric model topological relation, comprising: extracting the geometric feature of driving training vehicle to construct vehicle composite model;Extract the geometric feature of target position to construct target geometric model;Extract the outer contour polygon of each driving training ground to construct the driving school geometric model containing all driving training ground;According to the topological relation of the point geometric model of driving training vehicle and driving school geometric model, determine the driving training ground where driving training vehicle is located;Take the road geometric model and marker geometric model in the driving training ground where driving training vehicle is located to carry out collision judgment.The present application reaches higher accuracy within the accuracy range framework of existing positioning data, and is reliable, can provide timely and comprehensive collision information feedback.

Description

Technical Field

[0001] This invention belongs to the field of driver training technology, specifically relating to a method and device for dynamic collision detection between a driver training vehicle and a target location. Background Technology

[0002] Currently, the most commonly used collision detection algorithms in the driver training field are 3D rigid body detection and simple spatial geometric relationship methods. The first step is to create a 3D model of the site features, then place the modeled objects into a 3D game engine, using the engine's collision detection mechanism to determine if a collision has occurred. This method has two drawbacks: firstly, it requires a high degree of accuracy in the 3D modeling of the real-world scene and objects, is easily affected by environmental factors, and has high application costs; secondly, the process is complex, requiring the processing of a large amount of 3D data, and placing stringent demands on the processor's computing power.

[0003] The logic of the simple spatial geometric relationship method for collision detection commonly used in driver training is as follows: (1) Establish a two-dimensional spatial coordinate system, abstract the vehicle positioning points into point models, and abstract the site markings into circular models according to the teaching plan; (2) Determine whether a collision has occurred by the topological relationship between the car point coordinates and the circular model. In practice, this method is prone to missing key modeling steps, and because the modeling workload is huge, engineers are at a loss and cannot verify whether a collision has occurred. In addition, this method cannot collect comprehensive information and quantitative feedback on collision events, which is obviously not the best solution for driver training.

[0004] Currently, there is an urgent need for an efficient technical method in the field of driver training. The method should be economical in terms of rapid modeling; computationally efficient and low-complexity in terms of algorithm; and have high detection accuracy, strong reliability, and sensitive feedback in terms of teaching. Summary of the Invention

[0005] One of the objectives of this invention is to provide a dynamic collision detection method based on the topological relationship of a geometric model, which achieves high accuracy within the existing positioning data accuracy range, has strong reliability, and can provide timely and comprehensive collision information feedback.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A dynamic collision detection method based on geometric model topological relationships is applied to collision detection of driver training vehicles. The dynamic collision detection method based on geometric model topological relationships includes:

[0008] Extract the geometric features of the driver training vehicle to construct a vehicle composite model, which includes a point geometric model, a contour geometric model and a triangle geometric model;

[0009] Extract the geometric features of the target location to construct a target geometric model, which includes a road geometric model and a marker geometric model;

[0010] Extract the outer contour polygon of each driver training site to construct a driving school geometry model containing all driver training sites;

[0011] The location of the driving training vehicle is determined based on the topological relationship between the point geometric model of the driving training vehicle and the geometric model of the driving school.

[0012] Collision determination is performed using the road geometry model and marker geometry model of the driving training site where the driving training vehicle is located, including:

[0013] If the outline geometry model of the driver training vehicle has an intersecting topological relationship with the road geometry model in the driver training site, then it is determined that the driver training vehicle has collided with the road edge line in the driver training site.

[0014] Alternatively, if the triangular geometric model of the driver training vehicle has an intersecting topological relationship with the marked geometric model in the driver training area, then it is determined that the driver training vehicle has collided with the marked line in the driver training area, and the location of the collision point on the edge of the outline geometric model of the driver training vehicle is obtained, as well as the location of the collision point in the trisection of the marked line where the collision occurred.

[0015] Several alternative methods are provided below, but they are not intended as additional limitations on the overall solution above. They are merely further additions or optimizations. Provided there are no technical or logical contradictions, each alternative method can be combined individually with respect to the overall solution above, or multiple alternative methods can be combined with each other.

[0016] Preferably, the step of extracting the geometric features of the driver training vehicle to construct a vehicle composite model includes:

[0017] Extract the GNSS positioning points of the driver training vehicle as the point geometric model of the driver training vehicle;

[0018] Extract the outline polygon of the driver training vehicle as the outline geometric model of the driver training vehicle;

[0019] Extracting an equilateral or inverted triangle as the triangular geometric model for the driver training vehicle includes:

[0020] Using the GNSS positioning point of the driver training vehicle as the vertex of a triangle, with the GNSS positioning point located on the driver training vehicle, and the rear edge of the driver training vehicle as the base of the triangle, an equilateral triangle is formed in the lower half of the driver training vehicle, and the formed equilateral triangle is taken as the geometric model of the equilateral triangle; using the GNSS positioning point of the driver training vehicle as the vertex of a triangle, and the front edge of the driver training vehicle as the base of the triangle, an inverted triangle is formed in the upper half of the driver training vehicle, and the formed inverted triangle is taken as the geometric model of the inverted triangle.

[0021] Preferably, the step of extracting the geometric features of the target location to construct the target geometric model includes:

[0022] Use polylines or line segments to extract the road edge lines of the driver training site as the road geometry model for the target location;

[0023] The matching line segment method is used to extract the marking lines in the driver training site as the marking geometric model of the target location.

[0024] Preferably, determining the location of the driver training site based on the topological relationship between the point geometry model of the driver training vehicle and the driving school geometry model includes:

[0025] Take the outer contour polygon of a driver training area from the driving school's geometric model;

[0026] Take a point of the outer contour polygon as the judgment point, take the point geometric model from the judgment point to the driver training vehicle to form the first vector, take the next point of the outer contour polygon in a preset order to form the second vector, and calculate the cross product of the first vector and the second vector.

[0027] If the cross product values ​​calculated for each point on the outer contour polygon have the same sign, then it is determined that the driver training vehicle is within the driver training area.

[0028] Preferably, the topological relationship between the outline geometry model of the driver training vehicle and the road geometry model in the driver training site is determined, including:

[0029] The road geometry model is meshed;

[0030] The outline geometry model of the driver training vehicle is processed using a mesh and quadtree approach, and the processed outline data model is stored.

[0031] If any data point in the gridded data points corresponding to the road geometry model is contained in the contour data model, then it is determined that the contour geometry model of the driver training vehicle and the road geometry model in the driver training site have an intersecting topological relationship; otherwise, there is no intersecting topological relationship.

[0032] Preferably, determining the topological relationship between the triangular geometric model of the driver training vehicle and the marked geometric model in the driver training site includes:

[0033] The marked geometric model is meshed;

[0034] The triangular geometric model of the driver training vehicle is processed using a mesh and quadtree approach, and the processed triangular data model is stored.

[0035] If any data point in the gridded data point corresponding to the marked geometric model is contained in the triangular data model, then it is determined that the triangular geometric model of the driver training vehicle and the marked geometric model in the driver training site have an intersecting topological relationship; otherwise, there is no intersecting topological relationship.

[0036] Specifically, if the driver training vehicle is moving forward, the inverted triangle geometric model of the driver training vehicle is used to determine the topological relationship; if the driver training vehicle is moving backward, the equilateral triangle geometric model of the driver training vehicle is used to determine the topological relationship.

[0037] As a preferred option, if there are multiple marked geometric models in the driver training site, the topological relationship between each marked geometric model and the triangular geometric model of the driver training vehicle is determined.

[0038] Preferably, the dynamic collision detection method based on geometric model topological relationships further includes real-time correction, wherein the real-time correction includes:

[0039] The GNSS positioning points of the driver training vehicle within a preset time period are used to generate a real-time trajectory curve.

[0040] Two extreme trajectory lines parallel to the standard trajectory curve are generated based on the maximum range of steering wheel adjustment of the current driving training vehicle. The area between the two extreme trajectory lines is taken as the safe driving trajectory area.

[0041] The system determines whether the driver training vehicle is within the safe driving trajectory area based on the topological relationship between the point geometry model of the vehicle and the safe driving trajectory area. If the vehicle is within the safe driving trajectory area, no correction is prompted; otherwise, a correction is prompted.

[0042] This invention provides a dynamic collision detection method based on the topological relationships of geometric models. Combining the patterns of driver training sites and lesson plans, it extracts appropriate dynamic self-matching geometric models for tasks such as scene containment, edge collision, and target location. The topological relationships between these geometric models are then used to detect collision events between vehicles and target locations. A comprehensive, quantifiable, and visualized event description language is employed. Since this method only requires two-dimensional data for modeling, the modeling task is simple and cost-effective. The overall algorithm complexity is concise and efficient; for example, the complexity of scene containment is O(LogN). Within the existing accuracy range of positioning data, this method exhibits high accuracy and reliability, providing timely and comprehensive collision information feedback.

[0043] The second objective of this invention is to provide a dynamic collision detection device based on the topological relationship of a geometric model, which has high accuracy and reliability within the existing positioning data accuracy range and can provide timely and comprehensive collision information feedback.

[0044] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0045] A dynamic collision detection device based on geometric model topological relationships includes a processor and a memory storing a number of computer instructions, which, when executed by the processor, implement the steps of the dynamic collision detection method based on geometric model topological relationships. Attached Figure Description

[0046] Figure 1 This is a flowchart of the dynamic collision detection method based on the topological relationship of a geometric model according to the present invention;

[0047] Figure 2 This is an example diagram of the equilateral / inverted triangle geometric model in the vehicle composite model of the present invention;

[0048] Figure 3 This is an example diagram illustrating how the present invention determines whether point P is inside a polygon;

[0049] Figure 4 This is an example diagram of the geometric model of the marking lines in the driver training area of ​​this invention;

[0050] Figure 5 This is an example diagram of the octagonal model in the automotive composite model of the present invention;

[0051] Figure 6 This is an example diagram showing detailed information about the marker lines in the collision event of this invention;

[0052] Figure 7 This is an example diagram showing detailed information about the driver training vehicle in the collision event of this invention;

[0053] Figure 8 This is a schematic diagram of the limit trajectory of the vehicle's correct driving in a collision event according to the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention.

[0056] This invention addresses the shortcomings of existing collision detection algorithms, such as high modeling costs, high complexity of rigid body collision algorithms, insufficient detection accuracy, and difficulty in verification. It proposes a method based on constructing a geometric model using dynamic feature points and leveraging the dynamic processing of collision detection through the topological relationships of this geometric model. This method extracts multi-scene, multi-dimensional geometric models of the vehicle as a composite model. For target position characterization, a dynamic matching method is used to abstract the target model. The topological relationships of the geometric model are used to detect whether a collision occurs between the composite model and the target model. Furthermore, the relative positions of the two models at the time of a collision are recorded, and detailed and comprehensive information about the collision event is collected using methods such as angular deviation, providing timely feedback to the system with necessary alerts. Finally, real-time driving trajectories are collected, and geometric models of the test limit areas are acquired. The fit between the two models is analyzed, and real-time correction is provided.

[0057] like Figure 1 As shown in the figure, the dynamic collision detection method based on the topological relationship of the geometric model proposed in this embodiment includes the following steps:

[0058] S1. Extract the geometric features of the driver training vehicle to construct a composite vehicle model.

[0059] The geometric model that a vehicle needs to be abstracted or fitted varies depending on the scenario. For example, when determining the vehicle's location, it needs to be abstracted as a point; when determining whether the vehicle has collided with the roadside line, it needs to be abstracted as a vehicle outline polygon; when determining whether the vehicle has reached a target position, it needs to be abstracted as an equilateral or inverted triangle. Therefore, this step can be broken down into the following steps:

[0060] S1.1 Extract the high-precision GNSS positioning points of the vehicle in real time as the point geometric model C_Point of the vehicle.

[0061] S1.2 Extract the vehicle's outline polygons based on vehicle mapping (extract the vehicle outline polygons with the highest possible fit) as the outline geometric model C_Out.

[0062] S1.3 Extract an equilateral or inverted triangle as the triangular geometric model of the driver training vehicle. Using the real-time GNSS positioning point (located on the driver training vehicle) as the vertex of an isosceles triangle, obtain the inverted triangle geometric model C_Back and the equilateral triangle geometric model C_Front in the upper and lower halves of the vehicle, respectively. Figure 2 As shown.

[0063] In one embodiment, the method for extracting the equilateral triangle geometric model is as follows: using the vehicle's GNSS positioning point as the vertex of the triangle and the rear edge of the training vehicle as the base of the triangle, an equilateral triangle is formed in the lower half of the training vehicle, and the formed equilateral triangle is taken as the equilateral triangle geometric model. Considering the actual vehicle length, when forming the equilateral triangle, one can start from the vertex of the triangle, draw a central axis along the vehicle's rearward direction, with a central axis length of 86cm, and draw a base length of 1 meter along the vertical direction of the central axis, obtaining an isosceles triangle as the equilateral triangle geometric model.

[0064] Method for extracting the inverted triangle geometric model: Using the vehicle's GNSS positioning point as the vertex of the triangle and the front edge of the training vehicle as the base, an inverted triangle is formed in the upper half of the training vehicle. This inverted triangle is then used as the inverted triangle geometric model. Considering the actual vehicle length, when forming an equilateral triangle, one can start from the vertex of the triangle and draw a centerline along the vehicle's forward direction. The centerline is 86cm long. A base of 1 meter is drawn perpendicular to the centerline, resulting in an isosceles triangle as the inverted triangle geometric model.

[0065] S2. Extract the geometric features of the target location and construct the target geometric model.

[0066] There are two types of target locations: one is the road edge line, a naturally existing feature of the site, and the other is the marked location required by driver training. The feature extraction method for the first type is relatively simple, using polyline or line segment geometric models to extract edge line features; the extraction of the marked location for the second type adopts the matching line segment method.

[0067] S2.1. Extract the road geometry model T_S of the driving training site using polylines or line segments as the target location. A driving training site may contain one or more lane edges. In this embodiment, the road edges are the road lines within the driving training subject two site, and the road lines within the site belong to one and only to a specific driving training site.

[0068] S2.2. The matching line segment method is used to extract the marker lines in different areas of the driver training site as the marker geometric model of the target location. The marker lines are the features of the so-called target location in the driving school teaching plan, which are automatically matched by the specific scene. In this embodiment, it is abstracted into line segments of different lengths and angles as the geometric model T_L.

[0069] The matching line segment method (dynamic matching method) implemented in this embodiment matches a line segment with the vehicle's fixed-point satellite coordinates as the midpoint, the vehicle's direction as the perpendicular direction, and a length twice the vehicle's width. This line segment is recorded according to the above method based on the content of different driving training courses and is matched with the driving school and training ground.

[0070] Because the area near the parking space is relatively narrow, line segments longer than 2 meters are used as marker lines; anything longer would be meaningless. However, the area further away from the parking space is wider, and to prevent collisions and missed spots, the marker line length should be greater than 5 meters. The angle of the line segment is equal to the perpendicular angle of the tangent to the standard trajectory arc of the vehicle. The angle value of the marker line's placement position is equal to the perpendicular angle of the tangent to the standard trajectory line, and it continuously changes with the real-time trajectory line. Furthermore, the marker line also carries left, center, and right segment information; by comparing the trainee's actual trajectory line with the standard trajectory line, it determines in real time whether the trainee has traversed the correct path.

[0071] S3. Extract the outer contour polygon of each driving training site to construct a driving school geometry model SCE containing all driving training sites.

[0072] Generally, driving training grounds are irregular convex polygons, and the outline polygon can be enclosed by polygons with the fewest possible sides. Driving schools usually contain one or more driving training grounds, and more often than not, so the driving school geometric model in this embodiment contains one or more outer contour polygons.

[0073] S4. Determine the location of the driver training site based on the topological relationship between the point geometry model of the driver training vehicle and the geometry model of the driving school.

[0074] When performing collision detection, the first step is to determine whether a vehicle has entered or left one of the multiple driving school training areas. This involves determining the containment relationship between the point geometry model C_Point and the polygon SCE. This determination allows for real-time monitoring of whether driving school vehicles have changed training areas.

[0075] In the field of geometric topology, DE-9IM stands for Dimensionally Extended nine-Intersection Model, which has the following relationships: Contains, Covered By, Crosses, Disjoint, Equal, Inside, Intersects, Overlaps, Touch, and Within.

[0076] This embodiment uses the cross product rule to determine the topological relationship between the point geometric model of the driver training vehicle and the driving school geometric model. The cross product rule is as follows: if the cross product is greater than 0, then point 2(x2,y2) is to the left of point 1(x1,y1); if it is less than 0, then it is to the right; and if it is equal to 0, then the vectors are collinear.

[0077] Given the cross product of two two-dimensional vectors of two points: (x1, y1)*(x2, y2)=x1*y2-y1*x2;

[0078] If the value is greater than 0, it means that (x2,y2) is to the left of (x1,y1), and vice versa. If it is equal to 0, it means that the two vectors are collinear.

[0079] According to the rules above, the determination of whether a point C_Point in the geometric model is within the SCE polygon of the geometric model is as follows:

[0080] Take the outer contour polygon of a driver training area from the driving school's geometric model.

[0081] Take a point on the outer contour polygon as the judgment point, take the point geometric model from the judgment point to the driver training vehicle to form the first vector, take the next point on the outer contour polygon in a preset order to form the second vector, and calculate the cross product of the first vector and the second vector.

[0082] If the cross product values ​​calculated for each point on the outer contour polygon have the same sign, then it is determined that the driver training vehicle is within the driver training area.

[0083] Specifically, with Figure 3 The following example illustrates the determination of whether point P is inside polygon ABCDCE.

[0084] According to the cross product of vectors, if the cross products of vectors in counterclockwise (or clockwise) order have the same sign, then point P is inside the polygon. Specifically, the method is to take the cross products of vectors AB and AP, BC and BP, CD and CP, DE and DP, and EA and EP, and check if the resulting values ​​have the same sign. This helps determine the location of the point and whether it has left the area.

[0085] S5. Take the road geometry model and the marked geometry model in the driver training site where the driver training vehicle is located to make a collision judgment.

[0086] First, based on the index of the driver training site where the driver training vehicle is located obtained in step S4, the road geometry model T_S and the marker geometry models T_L1 to T_Ln in the driver training site are obtained.

[0087] The first type of collision is between the training vehicle and the road edge. In this case, the vehicle's composite model is a polygonal outline (C_Out), and the road edge is typically a polyline or line segment (T_S). The system determines whether there is a topological intersection between the two. If the outline geometry of the training vehicle and the road geometry in the training area intersect, then a collision occurs between the training vehicle and the road edge. The specific determination process is as follows:

[0088] a) Mesh the road geometry model T_S. If the angle of the line segment polyline is considered, use a meshing method with a quadtree to store the data model line.

[0089] b) Process the outline geometry model C_Out of the driver training vehicle using a meshing and quadtree approach, and store the processed outline data model polygon.

[0090] c) For all gridded data points contained in the line traversal data model (line) of the road geometry model T_S, determine whether the two intersect by checking if any point on the line is contained within the polygon of the contour data model. That is, if any data point in the gridded data model T_S is contained within the polygon of the contour data model, then the contour geometry model C_Out of the driver training vehicle and the road geometry model T_S in the driver training area have an intersecting topological relationship; otherwise, no intersecting topological relationship exists.

[0091] The second type of collision is between the training vehicle and the marked geometric model. In this case, the vehicle's composite model is a triangular geometric model. Combining the vehicle's driving information (OBD), if the vehicle is moving forward, an inverted triangular geometric model is used; if it is moving backward, an equilateral triangular geometric model is used. The same method as the intersection detection in the first case can be used to detect whether a marker line collision has occurred. The specific steps are as follows:

[0092] a) Mesh the marked geometric model;

[0093] b) Process the triangular geometric model of the driver training vehicle using a mesh and quadtree approach, and store the processed triangular data model;

[0094] c) If any data point in the gridded data points corresponding to the marked geometric model is contained within the triangular data model, then it is determined that the triangular geometric model of the driver training vehicle and the marked geometric model in the driver training area have an intersecting topological relationship; otherwise, no intersecting topological relationship exists. If there are multiple marked geometric models in the driver training area, then the topological relationship between each marked geometric model and the triangular geometric model of the driver training vehicle is determined one by one.

[0095] like Figure 4 As shown, the collision model for the marker lines in this embodiment uses segmented lines divided into three equal parts, such as... Figure 5 As shown, in this embodiment, an octagonal model of the driver training vehicle is obtained after a collision with the marking line, and detailed information in the collision event with the marking line is recorded based on this model.

[0096] In this embodiment, an octagonal model is used as the contour geometry model. In actual operation, there is no limit to the number of sides of the contour geometry model used. The octagon is a relatively simple vehicle shape to calculate. The more sides there are, the more accurate the collision point is, but the more computational burden there is. In the fitting method of the octagonal model, the four straight sides are based on the vehicle mapping, and the four hypotenuses (the curves where the four corners are located) are fitted by the lines connecting the two endpoints.

[0097] In this embodiment, when it is determined that a collision has occurred between a driver training vehicle and a marked line in the driver training area, the location of the collision point on the octagonal model of the driver training vehicle is obtained, as well as the location of the collision point in the trisection of the marked line where the collision occurred. The collision details recorded in this embodiment are used to collect comprehensive information on the collision time and provide quantitative feedback.

[0098] Based on the two types of collisions mentioned above, the specific steps for collision detection in this embodiment are as follows:

[0099] S5.1 When step S4 detects that a training vehicle has entered a training area SCE, each subsequent frame will determine in real time whether the training vehicle has left the training area. If the training vehicle leaves, the current step will be interrupted and the process will return to step S4 for loop detection; if it has not left, the next detection step will continue.

[0100] S5.2 In each frame, it is determined in real time whether the contour geometry model C_Out intersects with the road geometry model T_S in a topological relationship. If they intersect, the current step is interrupted and the process returns to step S4 for loop detection; if they do not intersect, the next detection step is continued.

[0101] S5.3 In each frame, it is determined in real time whether the geometric model C_Front / C_Back intersects with the marked geometric model T_L1 in a topological relationship. If there is no intersection, the current detection is repeated until an intersection occurs; if an intersection occurs, the detection step S5.3 is recursively performed. Note that when recursively performing the S5.3 step, T_L1 needs to be incremented by one to T_L2. The recursive call continues until T_Ln is reached, at which point the S5.3 step is exited, and the next detection step is performed.

[0102] The method for determining the intersection relationship between T_L and C_Front / C_Back in the steps is as follows: First, optimize the storage method of the data structure by performing gridding (Raster) on the data according to rows and columns, or by performing gridding and quadtree processing first, and storing it in a spatial data structure with optimized index time complexity. By traversing whether the gridded points contained in the line segment are in the polygon C_Front / C_Back, the intersection relationship between the line segment and the polygon is finally determined.

[0103] S5.4 Extract the detailed model of the marked geometric model in the driver training site, and determine the specific collision position and angle information between the vehicle and the model.

[0104] The extraction method for the refined model of the marker line in the steps is as follows: the line segment is divided into three parts: left (T_L_Left), middle (T_L_Middle), and right (T_L_Right). If the collision area of ​​S5.3 occurs in the middle part, it indicates that the passage is correct. If the collision occurs on the left or right, this deviation signal is fed back outward to provide timely reminders.

[0105] like Figure 6 As shown, the offset value of the angle information in the steps is obtained by the difference between the tangent angle of the vehicle trajectory arc when the actual collision occurs and the tangent angle of the standard trajectory arc, and this deviation signal is fed back to provide timely reminders.

[0106] S5.5 Extract detailed model information of the driver training vehicle and determine the specific collision location between the model and the T_L target marker line.

[0107] The method for extracting detailed models of driver training vehicles in the steps is as follows: Figure 7 As shown, the vehicle geometry model is extracted using a fitting method with a shape larger than an octagon. The collision point is detected to occur on one or more sides of the octagon, and the system then feeds back this signal.

[0108] Based on this, this embodiment achieves real-time collision detection of driver training vehicles and outputs comprehensive collision information, which has great application prospects in the driver training field. In another embodiment, based on collision detection, this embodiment also performs real-time correction to facilitate timely collision warning and reduce driver training losses. That is, this embodiment adds step S6 as follows, based on steps S1-S5 of the aforementioned embodiment.

[0109] S6. Extract the real-time trajectory curve geometric model, perform fitting data analysis with the standard trajectory curve geometric model, and make real-time correction and error summary analysis.

[0110] S6.1. Take the GNSS positioning points of the driver training vehicle within a certain period of time to form a real-time trajectory curve.

[0111] S6.2 Generate two extreme trajectory lines parallel to the standard trajectory curve based on the maximum range of steering wheel adjustment of the current driving training vehicle, and take the area between the two extreme trajectory lines as the safe driving trajectory area.

[0112] The standard trajectory curve in this embodiment uses the current driving school's head instructor as the driver and the training course provided by him as the operating method. The driver drives at a low speed (1-2 km / h) and must stop to change the steering wheel and direction of travel. This process is repeated 10 times. The curve formed by the GNSS fixed point of the driving training vehicle is the standard trajectory curve.

[0113] like Figure 8As shown, geometric feature extraction and modeling are performed for the safe zone where trainees pass through the prescribed route without retracing their steps. Since the vehicle angle and position are relatively fixed, the steering wheel angle that trainees can control during driving is known. Within a certain area, the dynamic changes in the steering wheel can be used to correct the entry into the parking space. However, the actual estimation at this time does not fit well with the standard trajectory curve. It is necessary to divide this trajectory area into several rectangular areas of about 10cm using a segmentation method. The limit range that the car can reach by steering wheel in the front and rear areas of each rectangular area is marked. Connecting these marked points, the upper and lower edges of the limit curve are derived. The upper and lower edges and the middle area together form the safe driving trajectory area, which is the safe driving trajectory area divided by the standard trajectory curve.

[0114] S6.3. Determine whether the driver training vehicle is within the safe driving trajectory area based on the topological relationship between the point geometry model of the driver training vehicle and the safe driving trajectory area. If it is within the safe driving trajectory area, do not prompt for correction; otherwise, prompt for correction.

[0115] The same algorithm is used in step S4 to determine the inclusion relationship between the vehicle point geometric model C_Point and the safe area of ​​the driving trajectory. If the vehicle point geometric model C_Point is within the safe area of ​​the driving trajectory, no warning is given; otherwise, a real-time correction warning is given.

[0116] This embodiment can also compare the fitting degree of the real-time trajectory curve of the driver training vehicle with the standard trajectory curve when the driving stage is completed, and analyze the cause of the error, such as turning the steering wheel too late or the speed being unbalanced. It should be noted that the focus of this embodiment is to provide the fitting degree analysis of the real-time trajectory curve and the standard trajectory curve. How to determine the cause of the error based on the analysis results is not the focus of this embodiment, so the specific error cause judgment logic will not be described in detail.

[0117] This embodiment first extracts a multi-scenario, multi-dimensional composite model of the vehicle, extracts the dynamic self-matching geometric model features of the target location in the driver training site, then uses the ray method to determine the site where the vehicle's point geometric model is located, and uses the mesh method or mesh method plus quadtree method to calculate the topological relationship between the aforementioned vehicle and target object geometric models, ultimately realizing collision detection and detailed description of collision events.

[0118] In another embodiment, this application also provides a dynamic collision detection device based on geometric model topological relationships, including a processor and a memory storing a plurality of computer instructions, wherein the computer instructions, when executed by the processor, implement the steps of the dynamic collision detection method based on geometric model topological relationships.

[0119] For specific limitations on dynamic collision detection devices based on geometric model topological relationships, please refer to the limitations on dynamic collision detection methods based on geometric model topological relationships mentioned above, which will not be repeated here.

[0120] The memory and processor are electrically connected directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected to each other via one or more communication buses or signal lines. The memory stores a computer program that can run on the processor, which implements the dynamic collision detection method based on geometric model topology relationships in this embodiment of the invention by running the computer program stored in the memory.

[0121] The memory may be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The memory stores the program, and the processor executes the program upon receiving an execution instruction.

[0122] The processor may be an integrated circuit chip with data processing capabilities. The aforementioned processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor.

[0123] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0124] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A dynamic collision detection method based on geometric model topological relationships, applied to collision detection of driver training vehicles, characterized in that, The dynamic collision detection method based on geometric model topological relationships includes: Extract the geometric features of the driver training vehicle to construct a vehicle composite model, which includes a point geometric model, a contour geometric model and a triangle geometric model; Extract the geometric features of the target location to construct a target geometric model, which includes a road geometric model and a marker geometric model; Extract the outer contour polygon of each driver training site to construct a driving school geometry model containing all driver training sites; The location of the driving training vehicle is determined based on the topological relationship between the point geometric model of the driving training vehicle and the geometric model of the driving school. Collision determination is performed using the road geometry model and marker geometry model of the driving training site where the driving training vehicle is located, including: If the outline geometry model of the driver training vehicle has an intersecting topological relationship with the road geometry model in the driver training site, then it is determined that the driver training vehicle has collided with the road edge line in the driver training site. Alternatively, if the triangular geometric model of the driver training vehicle has an intersecting topological relationship with the marked geometric model in the driver training area, then it is determined that the driver training vehicle has collided with the marked line in the driver training area, and the location of the collision point on the edge of the outline geometric model of the driver training vehicle is obtained, as well as the location of the collision point in the trisection of the marked line where the collision occurred.

2. The dynamic collision detection method based on geometric model topological relationships as described in claim 1, characterized in that, The process of extracting geometric features of driver training vehicles to construct a composite vehicle model includes: Extract the GNSS positioning points of the driver training vehicle as the point geometric model of the driver training vehicle; Extract the outline polygon of the driver training vehicle as the outline geometric model of the driver training vehicle; Extracting an equilateral or inverted triangle as the triangular geometric model for the driver training vehicle includes: Using the GNSS positioning point of the driver training vehicle as the vertex of a triangle, with the GNSS positioning point located on the driver training vehicle, and the rear edge of the driver training vehicle as the base of the triangle, an equilateral triangle is formed in the lower half of the driver training vehicle, and the formed equilateral triangle is taken as the geometric model of the equilateral triangle; using the GNSS positioning point of the driver training vehicle as the vertex of a triangle, and the front edge of the driver training vehicle as the base of the triangle, an inverted triangle is formed in the upper half of the driver training vehicle, and the formed inverted triangle is taken as the geometric model of the inverted triangle.

3. The dynamic collision detection method based on geometric model topological relationships as described in claim 1, characterized in that, The step of extracting the geometric features of the target location to construct the target geometric model includes: Use polylines or line segments to extract the road edge lines of the driver training site as the road geometry model for the target location; The matching line segment method is used to extract the marking lines in the driver training site as the marking geometric model of the target location.

4. The dynamic collision detection method based on geometric model topological relationships as described in claim 1, characterized in that, The step of determining the location of a driver training vehicle based on the topological relationship between the point geometry model of the vehicle and the driving school geometry model includes: Take the outer contour polygon of a driver training area from the driving school's geometric model; Take a point of the outer contour polygon as the judgment point, take the point geometric model from the judgment point to the driver training vehicle to form the first vector, take the next point of the outer contour polygon in a preset order to form the second vector, and calculate the cross product of the first vector and the second vector. If the cross product values ​​calculated for each point on the outer contour polygon have the same sign, then it is determined that the driver training vehicle is within the driver training area.

5. The dynamic collision detection method based on geometric model topological relationships as described in claim 1, characterized in that, Determining the topological relationship between the outline geometry model of the driver training vehicle and the road geometry model in the training ground includes: The road geometry model is meshed; The outline geometry model of the driver training vehicle is processed using a mesh and quadtree approach, and the processed outline data model is stored. If any data point in the gridded data points corresponding to the road geometry model is contained in the contour data model, then it is determined that the contour geometry model of the driver training vehicle and the road geometry model in the driver training site have an intersecting topological relationship; otherwise, there is no intersecting topological relationship.

6. The dynamic collision detection method based on geometric model topological relationships as described in claim 2, characterized in that, Determine the topological relationship between the triangular geometric model of the driver training vehicle and the marked geometric model in the driver training area, including: The marked geometric model is meshed; The triangular geometric model of the driver training vehicle is processed using a mesh and quadtree approach, and the processed triangular data model is stored. If any data point in the gridded data point corresponding to the marked geometric model is contained in the triangular data model, then it is determined that the triangular geometric model of the driver training vehicle and the marked geometric model in the driver training site have an intersecting topological relationship; otherwise, there is no intersecting topological relationship. Specifically, if the driver training vehicle is moving forward, the inverted triangle geometric model of the driver training vehicle is used to determine the topological relationship; if the driver training vehicle is moving backward, the equilateral triangle geometric model of the driver training vehicle is used to determine the topological relationship.

7. The dynamic collision detection method based on geometric model topological relationships as described in claim 6, characterized in that, If there are multiple marked geometric models in the driver training site, the topological relationship between each marked geometric model and the triangular geometric model of the driver training vehicle is determined.

8. The dynamic collision detection method based on geometric model topological relationships as described in claim 1, characterized in that, The dynamic collision detection method based on geometric model topological relationships further includes real-time correction, which includes: The GNSS positioning points of the driver training vehicle within a preset time period are used to generate a real-time trajectory curve. Two extreme trajectory lines parallel to the standard trajectory curve are generated based on the maximum range of steering wheel adjustment of the current driving training vehicle. The area between the two extreme trajectory lines is taken as the safe driving trajectory area. The system determines whether the driver training vehicle is within the safe driving trajectory area based on the topological relationship between the point geometry model of the vehicle and the safe driving trajectory area. If the vehicle is within the safe driving trajectory area, no correction is prompted; otherwise, a correction is prompted.

9. A dynamic collision detection device based on geometric model topological relationships, comprising a processor and a memory storing a plurality of computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the dynamic collision detection method based on the topological relationship of the geometric model as described in any one of claims 1 to 8.

Images