Target detection method and system based on vehicle infrastructure cooperation

Abstract

The invention relates to the technical field of vehicle-road cooperation, in particular to a target detection method and system based on vehicle-road cooperation. The method comprises the following steps: acquiring road end point cloud data and vehicle end point cloud data, wherein the road end point cloud data comprises historical road end point cloud data and real-time road end point cloud data; obtaining a background template point cloud by using the historical road end point cloud data, and obtaining a road side foreground point cloud by combining the real-time road end point cloud data and the background template point cloud; performing registration fusion on the vehicle end point cloud data and the roadside foreground point cloud to obtain target detection point cloud data; and completing target detection and identification based on the target detection point cloud data. According to the method, the data transmission quantity is reduced by optimizing the road end point cloud data, the target detection speed is improved, meanwhile, the point cloud data at the two ends of the vehicle and the road are effectively fused through the point cloud registration method, the target detection accuracy of the intelligent network connection vehicle is improved, and the driving safety is improved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle-road collaboration technology, and in particular to a target detection method and system based on vehicle-road collaboration. Background Art

[0002] With the continuous development and application of artificial intelligence, big data, and 5G technology, autonomous driving technology has ushered in rapid development. Environmental perception, as an important research topic of autonomous driving, is an important prerequisite for ensuring vehicle driving safety. The quality of environmental perception performance affects the path planning, decision-making, and control performance of autonomous driving. In the past period of time, researchers have focused on the intelligent environmental perception of single vehicles, and have continuously increased the number of on-board computing power and sensors to improve the target detection performance of autonomous driving vehicles; however, single-vehicle intelligence has problems such as blind spots and unstable perception at medium and long distances. The emergence of vehicle-road cooperative systems provides a new solution for autonomous driving environmental perception, which can well make up for the problems of single-vehicle intelligence. The vehicle-road cooperative system can obtain real-time traffic status and road condition information through communication with road infrastructure. With the perspective and performance advantages of roadside sensors, it assists vehicles in obtaining more comprehensive target detection perception results to ensure vehicle driving safety.

[0003] Target detection is a key task in the field of computer vision. It aims to accurately locate and identify target objects in images by analyzing sensors such as lidar and cameras. Applying vehicle-road collaboration technology to target detection will improve the performance of target detection algorithms under single-viewing angles. In summary, target detection based on vehicle-road collaboration is of great significance to the safety of unmanned driving. The quality of the vehicle-road collaboration target detection algorithm plays a key role in whether intelligent connected vehicles can cope with complex scenes at intersections. Therefore, it is very necessary to study the vehicle-road collaboration target detection algorithm to promote the development of intelligent connected vehicles. Summary of the invention

[0004] In view of the shortcomings of existing methods and the needs of practical applications, in order to improve the target detection capability of intelligent connected vehicles in complex scenarios such as intersections, and solve the blind spot and occlusion problems under a single perspective. On the one hand, the present invention provides a target detection method based on vehicle-road collaboration, comprising the following steps: Obtain road point cloud data and vehicle point cloud data, the road point cloud data includes historical road point cloud data and real-time road point cloud data; use the historical road point cloud data to obtain background template point cloud, combine the real-time road point cloud data and the background template point cloud to obtain roadside foreground point cloud; perform registration and fusion on the vehicle point cloud data and the roadside foreground point cloud to obtain target detection point cloud data; complete target detection and recognition based on the target detection point cloud data. The present invention optimizes the road point cloud data to reduce the amount of data transmission, which is beneficial to improving the target detection speed. At the same time, through the point cloud registration method, the point cloud data at both ends of the vehicle and the road are effectively integrated for target detection, thereby improving the accuracy of target detection of intelligent connected vehicles, solving the blind spot and occlusion problems under a single perspective, and helping to improve driving safety.

[0005] Optionally, the target detection method based on vehicle-road collaboration further includes the following steps: Construct a malicious node detection model; and use the malicious node detection model to eliminate malicious base station information. The present invention identifies malicious base stations through the malicious node detection model, effectively avoiding malicious node attacks, privacy leakage and identity tampering problems, and is conducive to improving the applicability of the present invention.

[0006] Optionally, the malicious node detection model satisfies the following formula: ,in, Represents the trust value, Indicates the number of successful interactions between a node and another node. Indicates the number of times a node fails to interact with another node. represents the trust threshold, , Represents the historical trust value between a node and another node, Represents the new trust value between a node and another node.

[0007] Optionally, combining the real-time road point cloud data and the background template point cloud to obtain a roadside foreground point cloud comprises the following steps: Traverse each point in the real-time road point cloud data to obtain the neighborhood corresponding to each point; construct a similarity model, and use the similarity model to obtain the similarity between each point and the background template point cloud; set a similarity threshold, and obtain the roadside foreground point cloud according to the similarity and the similarity threshold. The present invention sets the background template point cloud once, reduces the amount of later data matching, and then quickly obtains the point cloud data of the part of interest, which is conducive to improving the detection efficiency of the present invention.

[0008] Optionally, the similarity model satisfies the following formula:

[0009] in, Indicates The similarity between the point cloud and the background template point cloud, represents the length of the neighborhood, represents the width of the neighborhood, represents the height of the neighborhood, Indicates The coordinates of the point cloud on the X axis, Indicates The coordinates of the point cloud on the Y axis, Indicates The coordinates of the point cloud on the Z axis, Indicates The number of point clouds that belong to the background template point cloud are contained in the neighborhood of the point cloud.

[0010] Optionally, the registering and fusing the vehicle end point cloud data and the roadside foreground point cloud to obtain target detection point cloud data comprises the following steps: The vehicle endpoint cloud data and the roadside foreground point cloud are subjected to voxel downsampling, and surface reconstruction is performed based on the voxel downsampling result to obtain vehicle endpoint cloud reconstruction data and roadside foreground reconstructed point cloud; a coarse registration transformation matrix is ​​obtained based on the vehicle endpoint cloud reconstruction data and the roadside foreground reconstructed point cloud; the roadside foreground reconstructed point cloud is subjected to rigid body transformation using the coarse registration transformation matrix, and based on an iterative closest point algorithm, the transformation result and the vehicle endpoint cloud reconstruction data are combined to obtain a fine registration transformation matrix; the target detection point cloud data is obtained by combining the fine registration transformation matrix, the vehicle endpoint cloud data and the roadside foreground point cloud. The present invention uses the reconstructed point cloud data to perform two registrations to obtain a fine registration transformation matrix, and then completes the point cloud data fusion to obtain the target detection point cloud data.

[0011] Optionally, obtaining a coarse registration transformation matrix according to the vehicle end point cloud reconstruction data and the roadside foreground reconstructed point cloud comprises the following steps: Through any two adjacent graphics in the surface graphics, obtain the geometric relationship feature vector between the adjacent graphics; use the geometric relationship feature vector to obtain the comprehensive feature vector of the surface graphics; according to the comprehensive feature vector, obtain the initial matching graphics of the vehicle end point cloud reconstruction data and the roadside foreground reconstructed point cloud; evaluate the initial matching graphics to obtain the optimal rough matching graphics, and then use the optimal rough matching graphics to obtain the rough registration transformation matrix. The present invention obtains all geometric relationship feature vectors of the surface graphics and then obtains the comprehensive feature vector, takes the one that meets the conditions as the initial matching graphics, and finally evaluates the optimal initial matching graphics to obtain the rough registration transformation matrix, which is conducive to quickly completing the initial data screening and matching, and improving the detection efficiency of the present invention.

[0012] Optionally, the use of the geometric relationship feature vector to obtain a comprehensive feature vector of the surface figure satisfies the following formula: ,in, Represents the comprehensive feature vector of the graph within the feature calculation range, Indicates the geometric relationship feature vector of the graphics within the feature calculation + range, Indicates the number of adjacent graphics within the feature calculation range. Indicates the number of graphics within the feature calculation range. The geometric relationship feature vectors of adjacent figures, Indicates the number of graphics within the feature calculation range. The area of ​​adjacent shapes.

[0013] Optionally, the evaluating the initial matching graph to obtain an optimal rough matching graph satisfies the following formula:

[0014] in, represents the initial matching error coefficient of the graphics, represents the number of initial matching graphs, Expressed as The distance difference between the initial matching graphics after conversion, Expressed as For the initial matching graph, Represents the pre-set distance difference. Through the model formula, the evaluation results are specifically quantified, which is further conducive to accurate and fast point cloud matching and thus completing target detection.

[0015] In the second aspect, in order to be able to efficiently execute a target detection method based on vehicle-road collaboration provided by the present invention, the present invention also provides a target detection system based on vehicle-road collaboration, including a processor, an input device, an output device and a memory, wherein the processor, input device, output device and memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute a target detection method based on vehicle-road collaboration as described in the first aspect of the present invention. A target detection system based on vehicle-road collaboration of the present invention has a compact structure and stable performance, and can stably execute a target detection method based on vehicle-road collaboration provided by the present invention, further improving the overall applicability and practical application capabilities of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flow chart of a target detection method based on vehicle-road collaboration provided by an embodiment of the present invention; Figure 2A framework diagram of a target detection system based on vehicle-road collaboration provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of a target detection device based on vehicle-road collaboration provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are only for illustration and are not intended to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details do not need to be adopted to implement the present invention. In other examples, in order to avoid confusing the present invention, known circuits, software or methods are not specifically described.

[0018] Throughout the specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment of the present invention. Therefore, the phrases "in one embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or subcombination. In addition, it should be understood by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale.

[0019] See also Figure 1 In order to improve the target detection capability of intelligent connected vehicles in complex scenarios such as intersections and solve the blind spot and occlusion problems under a single perspective, the present invention provides a target detection method based on vehicle-road collaboration, such as Figure 1 As shown, the method comprises the following steps: S1. Obtain road endpoint cloud data and vehicle endpoint cloud data, wherein the road endpoint cloud data includes historical road endpoint cloud data and real-time road endpoint cloud data.

[0020] The road side in vehicle-road collaboration refers to the road infrastructure side, which is an important part of the vehicle-road collaboration system. The road side equipment is mainly responsible for communicating with the vehicle side equipment and collecting information about the road and its surrounding environment to achieve comprehensive perception and collaborative control of the traffic environment. Specifically, the road side equipment usually includes intelligent roadside units (RSU), cameras, millimeter wave radars, lidars and other sensing devices, as well as high-performance computing units required for data processing. These devices exchange and share information with vehicle side equipment (such as intelligent on-board units OBU) through wireless communication technology, thereby achieving seamless collaboration between vehicles and road infrastructure.

[0021] The vehicle side is the part of the vehicle-road cooperative system that is responsible for real-time processing of massive data on the vehicle side and fusion of multi-sensor data. It uses various sensors and smart devices, such as lidar, cameras, millimeter-wave radar, etc., to perceive the information of the vehicle's surrounding environment in real time, and transmit this information to the roadside side or the cloud through the communication network for further processing and analysis Road point cloud data mainly comes from sensors installed on road infrastructure, such as LiDAR, cameras, etc. These sensors can capture the three-dimensional information of the road and its surrounding environment in real time to form point cloud data; LiDAR measures the distance between the object and the sensor by emitting laser pulses and receiving reflected signals to generate three-dimensional point cloud data; Common equipment includes ground laser scanners (such as Leica, Trimble), etc. These devices are usually installed on both sides of the road or on viaducts to scan the road and its surrounding environment; Cameras capture images of the road and its surrounding environment, and then use computer vision algorithms to synthesize these images into a complete three-dimensional model to generate point cloud data. Common equipment includes high-definition cameras, stereo cameras, etc. These devices can be installed on road infrastructure such as traffic lights and street lights. In addition to LiDAR and cameras, other sensors such as ultrasonic sensors and infrared sensors can also be used to obtain road point cloud data. These sensors have their own characteristics, and appropriate sensors can be selected according to actual needs for combination.

[0022] The vehicle-point cloud data mainly comes from sensors installed on the vehicle, such as vehicle-mounted laser radar, cameras, etc. These sensors can capture the three-dimensional information of the vehicle's surrounding environment in real time, providing decision-making basis for autonomous vehicles. Vehicle-mounted laser radar measures the distance between the vehicle and the surrounding environment by emitting laser pulses and receiving reflected signals, thereby generating three-dimensional point cloud data; common equipment includes vehicle-mounted laser radar (such as Velodyne, Ouster), etc. These devices are usually installed on the top or front of the vehicle to scan the vehicle's surrounding environment; vehicle-mounted cameras capture images of the vehicle's surrounding environment, and then use computer vision algorithms to process and analyze these images to generate point cloud data. Common equipment includes high-definition cameras, stereo cameras, etc. These devices can be installed on the front, side and rear of the vehicle to obtain image information from different angles.

[0023] When acquiring point cloud data, you need to pay attention to the accuracy and resolution of the data. The higher the accuracy and resolution, the better the quality of the point cloud data, and the more it can meet the needs of applications such as autonomous driving. In applications such as autonomous driving, the real-time and reliability of point cloud data are crucial. Therefore, when selecting sensors and data processing algorithms, you need to ensure that they can meet the requirements of real-time and reliability. At the same time, you also need to comply with relevant privacy and security regulations to ensure that you do not infringe on the privacy and rights of others.

[0024] Furthermore, the road endpoint cloud data at the current detection moment is the real-time road endpoint cloud data, and the road endpoint cloud data stored in the database or cloud before the current detection moment is the historical road endpoint cloud data.

[0025] In other embodiments, after acquiring the road endpoint cloud data and the vehicle endpoint cloud data, the vehicle-road collaboration-based target detection method further includes the following steps: S11. Build a malicious node detection model.

[0026] Specifically, the malicious node detection model satisfies the following formula:

[0027] in, Represents the trust value, Indicates the number of successful interactions between a node and another node. Indicates the number of times a node fails to interact with another node. represents the trust threshold, , Represents the historical trust value between a node and another node, Indicates the new trust value between a node and another node. With each detection, the trust value between nodes is obtained based on the data interaction of each device as a port node, and the current trust value is used as the new trust value, and the previous data is the historical trust value.

[0028] S12: Utilize the malicious node detection model to eliminate malicious base station information.

[0029] Specifically, the trust value and the trust threshold between base stations are calculated by the malicious node detection model, and the corresponding base stations with trust values ​​less than the trust threshold are removed as malicious nodes.

[0030] S2. Utilize the historical road point cloud data to obtain a background template point cloud, and combine the real-time road point cloud data and the background template point cloud to obtain a roadside foreground point cloud.

[0031] There is a lot of redundancy between the point cloud frames collected by road-side equipment at the same location. The background parts such as buildings, trees, and ground in the surrounding environment show large-area overlap and highly similar statistical characteristics, while the differences between the point cloud frames happen to be the foreground point clouds of interest.

[0032] In an embodiment, the background template point cloud is obtained by using the historical road point cloud data to remove all foreground target point clouds of interest. The point cloud of a single frame is relatively sparse, and there may be visual blind spots caused by target occlusion, which makes it impossible to make a comprehensive description of the surrounding background environment. Therefore, multiple frames of point clouds can also be spliced ​​to construct a background template point cloud that describes the surrounding environment more comprehensively. It should be understood that the collection of template point clouds is selected during the period when the foreground targets are sparse. If necessary, the foreground points of the targets of interest (including static foreground targets) are manually circled and removed to obtain a pure background template point cloud.

[0033] Furthermore, the step S2 combines the real-time road point cloud data and the background template point cloud to obtain the roadside foreground point cloud, including the following steps: S21. Traverse each point in the real-time road endpoint cloud data to obtain a neighborhood corresponding to each point.

[0034] In this embodiment, the neighborhood is centered on the point in the real-time road endpoint cloud data. For long, For width, is a high cube.

[0035] S22, constructing a similarity model, and using the similarity model to obtain the similarity between each point and the background template point cloud.

[0036] Specifically, the similarity model satisfies the following formula: ,in, Indicates The similarity between the point cloud and the background template point cloud, represents the length of the neighborhood, represents the width of the neighborhood, represents the height of the neighborhood, Indicates The coordinates of the point cloud on the X axis, Indicates The coordinates of the point cloud on the Y axis, Indicates The coordinates of the point cloud on the Z axis, Indicates The number of point clouds that belong to the background template point cloud are contained in the neighborhood of the point cloud.

[0037] S23: Setting a similarity threshold, and obtaining a roadside foreground point cloud according to the similarity and the similarity threshold.

[0038] In the embodiment, the similarity threshold is set according to practical experience, and the similarity obtained by the similarity model is compared with the similarity threshold, and the point cloud with a similarity greater than the similarity threshold is classified as a background template point cloud, otherwise it is a roadside foreground point cloud.

[0039] It should be understood that there are still a small number of outliers in the preliminary discrimination results. Outliers are also called isolated points, which refer to noise points that are obviously deviated from the foreground point cloud cluster. Most of these outliers come from background point clouds or measurement noise, which not only affects the detection performance of subsequent algorithms, but also takes up a certain amount of storage space. Therefore, it is necessary to perform outlier filtering based on the preliminary search results to obtain a roadside foreground point cloud with smaller data volume and purer.

[0040] S3, registering and fusing the vehicle end point cloud data and the roadside foreground point cloud to obtain target detection point cloud data.

[0041] In an embodiment, the step S3 of registering and fusing the vehicle end point cloud data and the roadside foreground point cloud to obtain target detection point cloud data includes the following steps: S31, performing voxel downsampling on the vehicle endpoint cloud data and the roadside foreground point cloud, and performing surface reconstruction based on the voxel downsampling result to obtain vehicle endpoint cloud reconstruction data and roadside foreground reconstructed point cloud.

[0042] The principle of voxel downsampling is to divide the three-dimensional space where the point cloud data is located into multiple cubes of fixed size (called voxels), and then retain only one point in each voxel as a representative to replace all the points in the voxel.

[0043] Specifically, the voxel downsampled three-dimensional point cloud is projected onto a two-dimensional plane through a greedy projection method, and then a polygonal mesh is established using a recursive growth method, a split-and-merge method, or a point-by-point insertion method. In this embodiment, a triangular mesh is used. After the establishment is completed, the two-dimensional polygonal mesh is restored to a three-dimensional space according to the three-dimensional topological relationship between the point clouds to complete the surface reconstruction.

[0044] S32, obtaining a coarse registration transformation matrix according to the vehicle end point cloud reconstruction data and the roadside foreground reconstructed point cloud.

[0045] In an embodiment, obtaining a coarse registration transformation matrix according to the vehicle end point cloud reconstruction data and the roadside foreground reconstructed point cloud comprises the following steps: S321. Obtain a geometric relationship feature vector between any two adjacent graphics in the surface graphics.

[0046] Taking any spatial polygon as the central figure and defining the local coordinate system, the geometric relationship feature vectors of other spatial polygons collinear with the central figure are obtained. Furthermore, the geometric relationship feature vectors include the normal vector of the spatial polygon and the unit vector of the local coordinate system. , , The angle between represents the unit normal vector of the central figure, Represents the unit vector from the center point of the central figure to the vertex, Representation and , Perpendicular unit vector.

[0047] S322. Obtain a comprehensive feature vector of the surface graphic using the geometric relationship feature vector.

[0048] In an embodiment, the geometric relationship feature vector is used to obtain a comprehensive feature vector of the surface figure, which satisfies the following formula: ,in, Represents the comprehensive feature vector of the graph within the feature calculation range, Indicates the geometric relationship feature vector of the graphics within the feature calculation + range, Indicates the number of adjacent graphics within the feature calculation range. Indicates the number of graphics within the feature calculation range. The geometric relationship feature vectors of adjacent figures, Indicates the number of graphics within the feature calculation range. The area of ​​adjacent shapes.

[0049] S323. Obtaining an initial matching graph of the vehicle end point cloud reconstruction data and the roadside foreground point cloud reconstruction data based on the comprehensive feature vector.

[0050] In the embodiment, the distance between the comprehensive feature vectors is calculated, and when the distance between the two feature vectors is less than or equal to the matching threshold, they are preliminarily identified as a matching feature pair, that is, the initial matching graph of the vehicle end point cloud reconstruction data and the roadside foreground reconstructed point cloud is obtained. Furthermore, the distance between the comprehensive feature vectors can be Euclidean distance, Manhattan distance, Minkowski distance or Chebyshev distance.

[0051] S324, evaluating the initial matching graph to obtain an optimal rough matching graph, and then using the optimal rough matching graph to obtain the rough registration transformation matrix.

[0052] Specifically, the initial matching graph is evaluated to obtain an optimal rough matching graph that satisfies the following formula:

[0053] in, represents the initial matching error coefficient of the graphics, represents the number of initial matching graphs, Expressed as The distance difference between the initial matching graphics after conversion, Expressed as For the initial matching graph, Indicates the preset distance difference.

[0054] Furthermore, based on the obtained optimal coarse matching graph, a coarse registration transformation matrix is ​​obtained by using the SAC-IA method.

[0055] S33, using the coarse registration transformation matrix to perform a rigid body transformation on the roadside foreground reconstructed point cloud, and based on an iterative closest point algorithm, combining the transformation result and the vehicle end point cloud reconstruction data, to obtain a fine registration transformation matrix.

[0056] Specifically, the coarse registration transformation matrix is ​​used to perform a rigid body transformation on the roadside foreground reconstructed point cloud to obtain a transformed point cloud; then, the transformed point cloud is used as the initial origin point cloud, and the vehicle end point cloud reconstruction data is used as the initial target point cloud, and an iterative closest point algorithm is used for calculation to obtain a fine registration transformation matrix.

[0057] S34. Combining the precise registration transformation matrix, the vehicle end point cloud data and the roadside foreground point cloud, obtain the target detection point cloud data.

[0058] In an embodiment, the precise registration transformation matrix is ​​used to perform a rigid body transformation on the roadside foreground point cloud, and the transformed vehicle endpoint cloud data will be aligned with the roadside foreground point cloud data in a unified coordinate system; then the transformed roadside foreground point cloud data and the vehicle endpoint cloud data are fused to obtain the target detection point cloud data.

[0059] S4. Complete target detection and recognition based on the target detection point cloud data.

[0060] In an embodiment, a target detection model is constructed, and target detection and recognition are completed using the target detection model and the target detection point cloud data.

[0061] Furthermore, the target detection model includes PointNet, PointRCNN, and STD deep learning networks.

[0062] See also Figure 2In an embodiment, in order to be able to efficiently execute a target detection method based on vehicle-road collaboration provided by the present invention, the present invention also provides a target detection system based on vehicle-road collaboration, including: an input device, an output device, a processor, and a memory, wherein the input device, the output device, the processor, and the memory are interconnected, and the memory contains program instructions, and the program instructions are used for the steps of the target detection method based on vehicle-road collaboration. The target detection system based on vehicle-road collaboration of the present invention has a compact structure and stable performance, and can stably execute a target detection method based on vehicle-road collaboration of the present invention, further improving the overall applicability and practical application capabilities of the present invention.

[0063] In an embodiment, the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. An input device may be used to obtain data information. An output device may be used to output the results obtained by storing program instructions contained in a computer program in a memory provided by the present invention. The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory.

[0064] In yet another alternative embodiment, see Figure 3 In order to efficiently execute the target detection method based on vehicle-road collaboration provided by the present invention, this embodiment also provides a target detection device based on vehicle-road collaboration, such as Figure 3 As shown, including: The memory 10 is used to store computer programs; the processor 20 is used to execute computer programs to implement the above-mentioned target detection method based on vehicle-road collaboration. The memory 10, the processor 20, the communication interface 31 and the communication bus 32. The memory 10, the processor 20 and the communication interface 31 all communicate with each other through the communication bus 32.

[0065] In an embodiment, the memory 10 is used to store one or more program instructions. The memory 10 may store program instructions for implementing the following functions: Obtain road point cloud data and vehicle point cloud data, wherein the road point cloud data includes historical road point cloud data and real-time road point cloud data; use the historical road point cloud data to obtain a background template point cloud, and combine the real-time road point cloud data and the background template point cloud to obtain a roadside foreground point cloud; align and fuse the vehicle point cloud data with the roadside foreground point cloud to obtain target detection point cloud data; and complete target detection and recognition based on the target detection point cloud data.

[0066] In one possible implementation, the memory 10 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, and at least one application required for a function, etc.; the data storage area may store data created during use. In addition, the memory 10 may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include NVRAM. The memory stores an operating system and operating instructions, executable modules or data structures, or a subset thereof, or an extended set thereof, wherein the operating instructions may include various operating instructions for implementing various operations. The operating system may include various system programs for implementing various basic tasks and processing hardware-based tasks.

[0067] The processor 20 may be a central processing unit (CPU), an application specific integrated circuit, a digital signal processor, a field programmable gate array or other programmable logic device, and the processor 20 may be a microprocessor or any conventional processor, etc. The processor 20 may call a program stored in the memory 10. The communication interface 31 may be an interface of a communication module, and is used to connect with other devices or systems.

[0068] Of course, it should be noted that Figure 3 The structure shown does not constitute a limitation on the target detection device based on vehicle-road collaboration in this embodiment. In actual applications, the target detection device based on vehicle-road collaboration may include Figure 3 More or fewer components than shown, or combinations of certain components.

[0069] An embodiment also provides a storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the above-mentioned target detection method based on vehicle-road collaboration are implemented.

[0070] The storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0071] In summary, the present invention optimizes the point cloud data at both ends of the road to reduce the amount of data transmission, which is beneficial to improving the target detection speed. At the same time, through the point cloud registration method, the point cloud data at both ends of the vehicle and the road are effectively integrated for target detection, thereby improving the accuracy of target detection of intelligent connected vehicles, solving the blind spot and occlusion problems under a single perspective, and helping to improve driving safety.

[0072] Therefore, the present invention effectively overcomes various shortcomings in the prior art and has high industrial utilization value.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. 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 invention, and they should all be included in the scope recorded in the present invention.

Claims

1. A target detection method based on vehicle-road collaboration, characterized in that: The target detection method based on vehicle-road collaboration includes the following steps: Acquire road endpoint cloud data and vehicle endpoint cloud data, wherein the road endpoint cloud data includes historical road endpoint cloud data and real-time road endpoint cloud data; Using the historical road point cloud data to obtain a background template point cloud, combining the real-time road point cloud data and the background template point cloud to obtain a roadside foreground point cloud; Performing registration and fusion on the vehicle end point cloud data and the roadside foreground point cloud to obtain target detection point cloud data; Based on the target detection point cloud data, target detection and recognition are completed.

2. The target detection method based on vehicle-road collaboration according to claim 1 is characterized in that: The following steps are also included: Build a malicious node detection model; The malicious node detection model is used to eliminate malicious base station information.

3. The target detection method based on vehicle-road collaboration according to claim 2 is characterized in that: The malicious node detection model satisfies the following formula: ,in, Represents the trust value, Indicates the number of successful interactions between a node and another node. Indicates the number of times a node fails to interact with another node. represents the trust threshold, , Represents the historical trust value between a node and another node, Represents the new trust value between a node and another node.

4. The target detection method based on vehicle-road collaboration according to claim 1, characterized in that: The step of combining the real-time road point cloud data and the background template point cloud to obtain a roadside foreground point cloud comprises the following steps: Traversing each point in the real-time road endpoint cloud data to obtain a neighborhood corresponding to each point; Constructing a similarity model, and using the similarity model to obtain the similarity between each point and the background template point cloud; A similarity threshold is set, and a roadside foreground point cloud is obtained according to the similarity and the similarity threshold.

5. The target detection method based on vehicle-road collaboration according to claim 4 is characterized in that: The similarity model satisfies the following formula: ,in, Indicates The similarity between the point cloud and the background template point cloud, represents the length of the neighborhood, represents the width of the neighborhood, represents the height of the neighborhood, Indicates The coordinates of the point cloud on the X axis, Indicates The coordinates of the point cloud on the Y axis, Indicates The coordinates of the point cloud on the Z axis, Indicates The number of point clouds that belong to the background template point cloud are contained in the neighborhood of the point cloud.

6. The target detection method based on vehicle-road collaboration according to claim 1 is characterized in that: The step of registering and fusing the vehicle end point cloud data and the roadside foreground point cloud data to obtain target detection point cloud data includes the following steps: Performing voxel downsampling on the vehicle endpoint cloud data and the roadside foreground point cloud, and performing surface reconstruction based on the voxel downsampling result to obtain vehicle endpoint cloud reconstruction data and roadside foreground reconstructed point cloud; Obtaining a coarse registration transformation matrix according to the vehicle end point cloud reconstruction data and the roadside foreground reconstructed point cloud; Using the coarse registration transformation matrix, the roadside foreground reconstructed point cloud is rigidly transformed, and based on an iterative closest point algorithm, the transformation result and the vehicle end point cloud reconstruction data are combined to obtain a fine registration transformation matrix; The target detection point cloud data is obtained by combining the precise registration transformation matrix, the vehicle end point cloud data and the roadside foreground point cloud.

7. The target detection method based on vehicle-road collaboration according to claim 6 is characterized in that: The step of obtaining a coarse registration transformation matrix according to the vehicle end point cloud reconstruction data and the roadside foreground point cloud reconstruction data comprises the following steps: Obtaining a geometric relationship feature vector between any two adjacent graphics in the surface graphics; Using the geometric relationship feature vector, obtaining a comprehensive feature vector of the surface figure; According to the comprehensive feature vector, an initial matching graph of the vehicle end point cloud reconstruction data and the roadside foreground point cloud reconstruction data is obtained; The initial matching graph is evaluated to obtain an optimal coarse matching graph, and then the coarse registration transformation matrix is ​​obtained using the optimal coarse matching graph.

8. The target detection method based on vehicle-road collaboration according to claim 7 is characterized in that: The comprehensive feature vector of the surface figure is obtained by using the geometric relationship feature vector, which satisfies the following formula: ,in, Represents the comprehensive feature vector of the graph within the feature calculation range, Indicates the geometric relationship feature vector of the graphics within the feature calculation + range, Indicates the number of adjacent graphics within the feature calculation range. Indicates the number of graphics within the feature calculation range. The geometric relationship feature vectors of adjacent figures, Indicates the number of graphics within the feature calculation range. The area of ​​adjacent shapes.

9. The target detection method based on vehicle-road collaboration according to claim 7, characterized in that: The initial matching graph is evaluated to obtain an optimal rough matching graph that satisfies the following formula: ,in, represents the initial matching error coefficient of the graphics, represents the number of initial matching graphs, Expressed as The distance difference between the initial matching graphics after conversion, Expressed as For the initial matching graph, Indicates the preset distance difference.

10. A target detection system based on vehicle-road collaboration, characterized in that: The target detection system based on vehicle-road collaboration includes: an input device, an output device, a processor, and a memory. The input device, output device, processor, and memory are interconnected. The memory includes program instructions, and the program instructions are used to execute the target detection method based on vehicle-road collaboration according to any one of claims 1-9.

Images