Target detection and tracking method, system, device and medium based on laser point cloud

Abstract

The present invention discloses a target detection and tracking method, system, device, and medium based on laser point clouds. The method comprises the following steps: obtaining initial point cloud data from digital signals received by a laser radar, preprocessing the initial point cloud data to obtain first point cloud data, applying fast Euclidean clustering to the first point cloud data to obtain clustered point cloud data, obtaining a list of detection targets based on the clustered point cloud data, associating and matching existing trajectories with the detection targets, and estimating the target trajectory using a Kalman filter algorithm. In low-speed unmanned driving scenarios, the present invention not only ensures target tracking effectiveness but also reduces the algorithm's requirements for computing equipment, thereby reducing the equipment cost of low-speed unmanned driving and avoiding "performance waste."

Description

Technical Field

[0001] The present invention relates to the field of low-speed unmanned driving target perception technology, and in particular to a target detection and tracking method, system, equipment and medium based on laser point cloud. Background Art

[0002] Object tracking technology was first applied in computer vision to track objects that appear continuously within a single video. With the development of laser point cloud technology, object tracking has gradually been applied to point cloud objects. Depending on the number of objects being tracked, object tracking can be categorized as single object tracking (SOT) or multiple object tracking (MOT). In the autonomous driving field, multi-object tracking technology is of particular interest.

[0003] There are two paradigms for multi-target tracking: the tracking-by-detection (TBD) paradigm and the joint-detection-and-tracking (JDT) paradigm. The TBD paradigm, as the name suggests, involves tracking by target detection. It first uses a target detection algorithm to detect the position and size of all targets in each frame of data; it then connects the detection results from different frames through data management. The core idea of ​​the JDT paradigm is to simultaneously implement target detection and tracking within a unified framework. It is an end-to-end trainable target tracking paradigm. Compared to the JDT paradigm, the TBD paradigm is more classic and commonly used in the autonomous driving field.

[0004] Currently, in common autonomous driving frameworks, target tracking methods based on laser point clouds often use deep learning detectors (such as PointPillars and CenterPoint). Deep learning detectors can better consider target categories. However, the use of deep learning detectors brings another problem: the increased computing requirements. Deep learning detectors often require expensive, high-performance computing devices with GPUs to achieve near-real-time performance. From a cost-effective perspective, achieving real-time target detection and tracking on low-performance computing devices is undoubtedly a major challenge for autonomous vehicle manufacturers.

[0005] The disclosure of the above background technology content is only used to assist in understanding the inventive concept and technical solution of the present invention. It does not necessarily belong to the prior art of this patent application, nor does it necessarily provide technical guidance. In the absence of clear evidence that the above content has been disclosed before the filing date of this patent application, the above background technology should not be used to evaluate the novelty and creativity of this application. Summary of the Invention

[0006] The purpose of the present invention is to provide a target detection and tracking method, which enables the algorithm to achieve tracking effects equivalent to those in a high-performance computing environment under low-cost and low-performance computing equipment.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] A real-time target detection and tracking method based on laser point cloud, including

[0009] Obtain initial point cloud data through the digital signal received by the laser radar;

[0010] Preprocessing the initial point cloud data to obtain first point cloud data;

[0011] Applying fast Euclidean clustering to the first point cloud data to obtain clustered point cloud data;

[0012] Obtaining a detection target list according to the clustered point cloud data;

[0013] Associating and matching the existing trajectory with the detection target;

[0014] The Kalman filter algorithm is used to estimate the target trajectory.

[0015] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions,

[0016] The fast Euclidean clustering further includes the following steps:

[0017] constructing a balanced binary tree for the first point cloud data to accelerate the search;

[0018] Create a labels data group and a removed data group that are the same size as the first point cloud data;

[0019] Create an empty adjacency list G;

[0020] Traversing all points in the first point cloud data, merging the label numbers of the points according to the adjacency table G, assigning the same label number information to the points with connected label numbers in the adjacency table G, and outputting the label number array of the first point cloud data;

[0021] The clustered point cloud data is obtained according to the tag number array.

[0022] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, traversing all points in the point cloud data includes:

[0023] If the value of the removed data group corresponding to the current point p is true, the current point p has been processed and is skipped;

[0024] Create a new label label p , label p Add to the adjacency list G;

[0025] Take the current point p as the center and r as the radius to search the radius and obtain the list of points adjacent to p; loop through the list of adjacent points and extract point q: when the label number of point q is label q Valid and not equal to the label number label of the point p p When label is created in the adjacency table G q and label p connection relationship; set the label number of the point q to label p , and set the removed value of the point q to true.

[0026] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the method further includes: using a directional envelope box to describe the detection target;

[0027] The existing trajectory and the detected target are associated and matched using the weighted bipartite graph optimal matching algorithm.

[0028] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the method of associating and matching the existing trajectory with the detection target using the weighted bipartite graph optimal matching algorithm specifically includes:

[0029] All detected targets in the latest frame of the current trajectory are defined as a set U, with a set size of m;

[0030] Define the current detection target as a set V, with a set size of n;

[0031] The similarity between the set U targets and the set V targets is calculated based on the weighted information of the center point, direction, and OBB intersection ratio to form a similarity matrix M. m×n ;

[0032] According to the similarity matrix M m×n Find the optimal target assignment in the set U and the protected set V, and the unassigned targets in the set U are regarded as disappeared targets, and the unassigned targets in the set V are regarded as newly added targets.

[0033] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the pre-processing of the initial point cloud data to obtain the first point cloud data includes:

[0034] Performing voxel filtering on the initial point cloud data, filtering the initial point cloud data using the voxel filtering; and / or,

[0035] Perform ground filtering on the point cloud data that has undergone the voxel filtering to obtain first point cloud data.

[0036] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, ground filtering is performed on the point cloud data after the voxel filtering, specifically including:

[0037] Dividing the point cloud data after the voxel filtering into 0.1m×0.1m grids on the xy plane, and independently performing height estimation on each grid, wherein the height estimation includes a maximum height and a minimum height;

[0038] Assume that the ideal ground height is h0, and the grids that meet the following two conditions are filtered out as ground:

[0039] The maximum and minimum heights of the grid are both within the range [h0-Δ1,h0+Δ1];

[0040] The difference between the maximum height and the minimum height of the grid is less than Δ2.

[0041] Among them, h0 is the ideal ground height, Δ1 is the height difference threshold, and Δ2 is the maximum and minimum height difference threshold of the grid.

[0042] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the target trajectory is estimated by using the Kalman filter algorithm, including:

[0043] Based on the optimal state estimation of the target at the previous moment and combined with the state transition equation of the system, the state at the current moment is predicted.

[0044] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the method of estimating the target trajectory using the Kalman filter algorithm further includes:

[0045] According to the estimation of the position and speed of the target at time k-1, combined with the state transfer equation Predict the state at the current moment k;

[0046] in, is the estimated state at time k-1, is the state at the current moment k, u k-1 is the external input, A and B are the coefficient matrices of the shape equation.

[0047] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the method of estimating the target trajectory using the Kalman filter algorithm further includes:

[0048] Predict the error covariance matrix at the current moment k;

[0049] The error covariance matrix equation is

[0050] in, is the prediction error covariance matrix at the current moment k, P k-1 is the error covariance matrix at time k-1, Q is the noise of the prediction model, A T is the transposed matrix of A.

[0051] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the method of estimating the target trajectory using the Kalman filter algorithm further includes:

[0052] According to the error covariance matrix of the prediction results The Kalman gain is calculated by the error covariance matrix R of the observation results. The calculation formula is

[0053] The observation results and prediction results are weighted and summed according to the Kalman gain to obtain the optimal state estimate at the current moment k, which is calculated as follows:

[0054] Among them, K k is the Kalman gain, H is the observation matrix, H T is the transposed matrix of H, z k is the target observation result.

[0055] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the method of estimating the target trajectory using the Kalman filter algorithm further includes:

[0056] Update the error covariance matrix P at the current moment k k , and its calculation formula is

[0057] According to another aspect of the present invention, a real-time target detection and tracking system based on laser point cloud is provided, comprising the following modules:

[0058] A point cloud data acquisition module is used to acquire initial point cloud data through digital signals received by the laser radar;

[0059] A preprocessing module preprocesses the initial point cloud data to obtain first point cloud data;

[0060] Clustering module, used to perform fast Euclidean clustering on the ground filtered point cloud data to obtain clustered point cloud data

[0061] A detection target list acquisition module, configured to acquire a detection target list based on the clustered point cloud data;

[0062] A matching module, which associates and matches the existing trajectory with the detection target;

[0063] The target trajectory acquisition module uses the Kalman filter algorithm to estimate the target trajectory.

[0064] According to another aspect of the present invention, the present invention provides an electronic device, comprising:

[0065] Memory for storing computer programs;

[0066] The processor is configured to execute the steps of any one of the above-mentioned methods for real-time target detection and tracking based on laser point cloud when executing the computer program stored in the memory.

[0067] According to another aspect of the present invention, the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it performs the steps of any of the above-mentioned real-time target detection and tracking methods based on laser point clouds.

[0068] The beneficial effects brought about by the technical solution provided by the present invention are as follows:

[0069] a. Use the Fast Euclidean Cluster (FEC) method for target tracking detection, replacing deep learning detectors for target tracking detection, while achieving tracking results and reducing "performance waste";

[0070] b. Fast Euclidean clustering, combined with point cloud preprocessing techniques such as voxel filtering and ground filtering, reduces the amount of point cloud data required for target detection. This improves the data level for real-time target tracking, enabling the algorithm to achieve tracking results comparable to those in high-performance computing environments on low-cost, low-performance computing devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0072] Figure 1 A schematic diagram of a target detection and tracking method provided by an exemplary embodiment of the present invention;

[0073] Figure 2 A schematic diagram of a target detection and tracking system provided as an exemplary embodiment of the present invention;

[0074] Figure 3 A flowchart of fast Euclidean clustering is provided for an exemplary embodiment of the present invention;

[0075] Figure 4 A schematic diagram of an electronic device is provided for an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0076] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0077] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0078] The flowcharts used in this application illustrate operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowcharts may not be implemented in order, and steps that have no logical contextual relationship may be reversed or performed simultaneously. In addition, those skilled in the art, guided by the contents of this application, may add one or more other operations to the flowcharts, or may remove one or more operations from the flowcharts.

[0079] The embodiments described herein are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the accompanying drawings can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application.

[0080] In one embodiment of the present invention, a real-time target detection and tracking method based on laser point cloud is provided. Figure 1 As shown, the following steps are included:

[0081] Step 1: Obtain initial point cloud data through the digital signal received by the laser radar;

[0082] Step 2: Preprocessing the initial point cloud data to obtain first point cloud data;

[0083] Step 3: Apply fast Euclidean clustering to the first point cloud data to obtain clustered point cloud data;

[0084] Step 4: Obtain a detection target list based on the clustered point cloud data;

[0085] Step 5: Associating and matching the existing trajectory with the detected target;

[0086] Step 6: Use the Kalman filter algorithm to estimate the target trajectory.

[0087] The present invention uses the Fast Euclidean Cluster (FEC) method as the target tracking detection method, replacing the mainstream deep learning detector as the target tracking detection method. In the low-speed unmanned driving scenario, it not only ensures the target tracking effect, but also reduces the algorithm's requirements for computing equipment, thereby reducing the equipment cost of low-speed unmanned driving and avoiding "performance waste".

[0088] In one embodiment of the present invention, the fast Euclidean clustering step in step 3 is as follows: Figure 3 As shown:

[0089] Step 31: Construct a KD tree from the point cloud data to speed up the search;

[0090] Step 32: Create a labels data group and a removed data group that are the same size as the point cloud data;

[0091] Step 33: Create an empty adjacency list G;

[0092] Step 34: Traverse all points in the point cloud data;

[0093] Step 35: merging the label numbers of the points according to the adjacency table G, assigning the same label number information to the points with connected label numbers in the adjacency table G, and outputting the label number array of the first point cloud data;

[0094] Step 36: Obtain clustered point cloud data according to the tag number array. The nearest neighbor query algorithm based on KD-Tree is an important preprocessing method to accelerate the Euclidean clustering algorithm. The fast Euclidean clustering process is as follows: Figure 2shown.

[0095] In one embodiment of the present invention, step 34 is specifically as follows: if the value of the removed data group corresponding to the current point p is true, the current point p has been processed, and the process is skipped;

[0096] Create a new label label p , label p Add to the adjacency list G;

[0097] Take the current point p as the center and r as the radius to search the radius and obtain the list of points adjacent to p; loop through the list of adjacent points and extract point q: when the label number of point q is label q Valid and not equal to the label number label of the point p p When label is created in the adjacency table G q and label p connection relationship; set the label number of the point q to label p , and set the removed value of the point q to true.

[0098] In one embodiment of the present invention, step 4 uses a directional envelope box to describe the detection target.

[0099] In one embodiment of the present invention, step 5 uses a weighted bipartite graph optimal matching algorithm to associate and match the existing trajectory with the detection target.

[0100] In one embodiment of the present invention, the existing trajectory and the detection target are associated and matched using a weighted bipartite graph optimal matching algorithm, specifically including:

[0101] All detected targets in the latest frame of the current trajectory are defined as a set U, with a set size of m;

[0102] Define the current detection target as a set V, with a set size of n;

[0103] The similarity between the set U targets and the set V targets is calculated based on the weighted information of the center point, direction, and OBB intersection ratio to form a similarity matrix M. m×n ;

[0104] According to the similarity matrix M m×n Find the optimal target assignment in the set U and the protected set V, and the unassigned targets in the set U are regarded as disappeared targets, and the unassigned targets in the set V are regarded as newly added targets.

[0105] In one embodiment of the present invention, step 2 pre-processes the initial point cloud data, specifically comprising: performing voxel filtering on the initial point cloud data, filtering the initial point cloud data using the voxel filtering; and / or,

[0106] Perform ground filtering on the point cloud data that has undergone the voxel filtering to obtain first point cloud data.

[0107] Initial point cloud data has the following problems: (1) irregular point cloud data density needs to be smoothed; (2) outliers caused by occlusion and other issues need to be removed; (3) a large amount of data needs to be downsampled; (4) noise data needs to be removed. This paper uses point cloud preprocessing technologies such as voxel filtering and ground filtering to reduce the amount of point cloud data in the target detection process, making improvements at the data level to achieve real-time target tracking.

[0108] In addition, target detection is more important for sensing structures, moving objects, and people above the road surface. A large number of ground points will affect subsequent clustering, so it is preferable to filter out ground points. Ground filtering can be used to filter out ground points, thereby reducing the interference of ground points on target detection.

[0109] Furthermore, voxel filtering is used to filter the point cloud, and the voxel size is set to 0.01m. Then, ground filtering is performed, and the point cloud data is divided into 0.1m×0.1m grids on the xy plane. Each grid is independently estimated in height, and the height estimation includes the maximum height and the minimum height. Let the ideal ground height be h0, and the grids that meet the following two conditions are used as ground filters. The maximum height and minimum height of the grid are both in the range of [h0-Δ1,h0+Δ1], and the difference between the maximum height and the minimum height of the grid is less than Δ2, where h0 is the ideal ground height, Δ1 is the height difference threshold, and Δ2 is the maximum and minimum height difference threshold of the grid.

[0110] In one embodiment of the present invention, step 6 includes: Step 61: Based on the estimated position and velocity of the target at time k-1, combined with the state transfer equation Predict the state at the current moment k; where, is the estimated state at time k-1, is the state at the current moment k, u k-1 is the external input, A and B are the shape equation coefficient matrices; Step 62: predict the error covariance matrix at the current moment k to represent the uncertainty of the prediction; the error covariance matrix equation is in, is the prediction error covariance matrix at the current moment k, P k-1 is the error covariance matrix at time k-1, Q is the noise of the prediction model, A T is the transposed matrix of A;

[0111] Step 63: According to the error covariance matrix P of the prediction result k - The Kalman gain is calculated by the error covariance matrix R of the observation results. The calculation formula is

[0112] The observation results and prediction results are weighted and summed according to the Kalman gain to obtain the optimal state estimate at the current moment k, which is calculated as follows: Among them, K k is the Kalman gain, H is the observation matrix, H T is the transposed matrix of H, z k is the target observation result.

[0113] Step 64: Update the error covariance matrix P at the current time k k , and its calculation formula is

[0114] In one embodiment of the present invention, a real-time target detection and tracking system based on laser point cloud is provided. Figure 2 Shown, including:

[0115] A point cloud data acquisition module is used to acquire initial point cloud data through digital signals received by the laser radar;

[0116] A preprocessing module preprocesses the initial point cloud data to obtain first point cloud data;

[0117] Clustering module, used to perform fast Euclidean clustering on the ground filtered point cloud data to obtain clustered point cloud data

[0118] A detection target list acquisition module, configured to acquire a detection target list based on the clustered point cloud data;

[0119] A matching module, which associates and matches the existing trajectory with the detection target;

[0120] The target trajectory acquisition module uses the Kalman filter algorithm to estimate the target trajectory.

[0121] In one embodiment of the present invention, an electronic device is provided, such as Figure 4 As shown, the electronic device includes:

[0122] At least one processor 401, and a memory 402 connected to the at least one processor 401. The specific connection medium between the processor 401 and the memory 402 is not limited in the embodiment of the present application. Figure 4 In the example, the processor 401 and the memory 402 are connected via a bus 400. Figure 4The bus 400 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 The use of only one thick line in the figure does not indicate that there is only one bus or only one type of bus. Alternatively, processor 401 may also be referred to as a controller, without limitation to the name. In one embodiment of the present application, a computer-readable storage medium is provided, storing a computer program. When executed by a processor, the computer program implements the aforementioned network intrusion detection method.

[0123] In the embodiment of the present application, the memory 402 stores instructions that can be executed by at least one processor 401. The at least one processor 401 can execute the network intrusion detection method described above by executing the instructions stored in the memory 402. The processor 401 can implement Figure 4 The functions of each module in the device shown.

[0124] The processor 401 is the control center of the device, and can use various interfaces and lines to connect the various parts of the entire control device. By running or executing instructions stored in the memory 402 and calling data stored in the memory 402, the various functions of the device and processing data, the device can be monitored as a whole. In one possible design, the processor 401 may include one or more processing units. The processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 401. In some embodiments, the processor 401 and the memory 402 may be implemented on the same chip. In some embodiments, they may also be implemented on separate chips.

[0125] The processor 401 can be a general-purpose processor, such as a central processing unit (CPU), a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component, and can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the network intrusion detection method disclosed in the embodiments of the present application can be directly embodied as being executed by a hardware processor, or can be executed by a combination of hardware and software modules in the processor.

[0126] The memory 402 is a non-volatile computer-readable storage medium that can be used to store non-volatile software programs, non-volatile computer executable programs and modules. The memory 402 may include at least one type of storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory, a random access memory (Random Access Memory, RAM), a static random access memory (Static Random Access Memory, SRAM), a programmable read-only memory (Programmable Read Only Memory, PROM), a read-only memory (Read Only Memory, ROM), an electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), a magnetic memory, a disk, an optical disk, etc. The memory 402 is any other medium that can be used to carry or store a desired program code in the form of an instruction or data structure and can be accessed by a computer, but is not limited thereto. The memory 402 in the embodiment of the present application can also be a circuit or any other device that can realize a storage function, for storing program instructions and / or data.

[0127] By designing and programming the processor 401, the code corresponding to the network intrusion detection method in the above embodiment can be fixed into the chip, so that the chip can execute the code when running. Figure 1 The steps of the network intrusion detection method of the embodiment shown are as follows: How to design and program the processor 401 is a technique well known to those skilled in the art and will not be described in detail here.

[0128] In one embodiment of the present invention, a computer-readable storage medium is provided, wherein a computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, the steps of any one of the above-mentioned methods for real-time target detection and tracking based on laser point clouds are performed.

[0129] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0130] The above is only a specific implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A real-time target detection and tracking method based on laser point cloud, characterized in that: The following steps are involved: Obtain initial point cloud data through the digital signal received by the laser radar; The initial point cloud data is preprocessed, including: performing voxel filtering on the initial point cloud data, and filtering the initial point cloud data using the voxel filtering; performing ground filtering on the point cloud data after the voxel filtering to obtain first point cloud data, wherein the ground filtering includes: dividing the point cloud data after the voxel filtering into grids on the xy plane, and independently performing height estimation on each grid, and the height estimation includes a maximum height and a minimum height; assuming that the ideal ground height is h0, grids that meet the following two conditions are removed as ground filters: the maximum height and the minimum height of the grid are both within the range of [h0-Δ1, h0+Δ1]; the difference between the maximum height and the minimum height of the grid is less than Δ2, wherein h0 is the ideal ground height, Δ1 is the height difference threshold, and Δ2 is the maximum height and minimum height difference threshold of the grid; Fast Euclidean clustering is performed on the first point cloud data to obtain clustered point cloud data; the fast Euclidean clustering includes the following steps: constructing a balanced binary tree for the first point cloud data to accelerate the search; creating a labels data group and a removed data group of the same size as the first point cloud data; creating an empty adjacency list G; traversing all points in the first point cloud data, merging the label numbers of the points according to the adjacency list G, assigning the same label number information to the points with connected label numbers in the adjacency list G, and outputting the label number array of the first point cloud data; obtaining the clustered point cloud data according to the label number array; wherein, traversing all points in the point cloud data includes: if the value of the removed data group corresponding to the current point p is true, the current point p has been processed and is skipped; creating a new label label p , label p Add to the adjacency list G; take the current point p as the center and r as the radius to perform a radius search to obtain a list of points adjacent to p; loop through the list of adjacent points to extract point q: when the label number of point q is label q Valid and not equal to the label number label of the point p p When label is created in the adjacency table G q and label p connection relationship; set the label number of the point q to label p , and set the removed value of the point q to true; Obtaining a detection target list according to the clustered point cloud data; Associating and matching the existing trajectory with the detection target; The Kalman filter algorithm is used to estimate the target trajectory.

2. The real-time target detection and tracking method based on laser point cloud according to claim 1, characterized in that: Also includes: Using a directional envelope box to describe the detection target; The existing trajectory and the detected target are associated and matched using the weighted bipartite graph optimal matching algorithm.

3. The real-time target detection and tracking method based on laser point cloud according to claim 2, characterized in that: The method of associating and matching the existing trajectory with the detection target using the weighted bipartite graph optimal matching algorithm specifically includes: All detected targets in the latest frame of the current trajectory are defined as a set U, with a set size of m; Define the current detection target as a set V, with a set size of n; The similarity between the set U targets and the set V targets is calculated based on the weighted information of the center point, direction, and OBB intersection ratio to form a similarity matrix M. m×n ; According to the similarity matrix M m×n Find the optimal target assignment in the set U and the protected set V, and the unassigned targets in the set U are regarded as disappeared targets, and the unassigned targets in the set V are regarded as newly added targets.

4. The real-time target detection and tracking method based on laser point cloud according to claim 1, characterized in that: The target trajectory is estimated by using the Kalman filter algorithm, including: Based on the optimal state estimation of the target at the previous moment and combined with the state transition equation of the system, the state at the current moment is predicted.

5. The real-time target detection and tracking method based on laser point cloud according to claim 4, characterized in that: The method of estimating the target trajectory by using the Kalman filter algorithm further includes: According to the estimation of the position and speed of the target at time k-1, combined with the state transfer equation Predict the state at the current moment k; in, is the estimated state at time k-1, is the state at the current moment k, u k-1 is the external input, A and B are the coefficient matrices of the shape equation.

6. The real-time target detection and tracking method based on laser point cloud according to claim 4 or 5, characterized in that: The method of estimating the target trajectory by using the Kalman filter algorithm further includes: Predict the error covariance matrix at the current moment k; The error covariance matrix equation is in, is the prediction error covariance matrix at the current moment k, P k-1 is the error covariance matrix at time k-1, Q is the noise of the prediction model, A T is the transposed matrix of A.

7. The real-time target detection and tracking method based on laser point cloud according to claim 6, characterized in that: The method of estimating the target trajectory by using the Kalman filter algorithm further includes: According to the error covariance matrix of the prediction results The Kalman gain is calculated by the error covariance matrix R of the observation results. The calculation formula is The observation results and prediction results are weighted and summed according to the Kalman gain to obtain the optimal state estimate at the current moment k, which is calculated as follows: Among them, K k is the Kalman gain, H is the observation matrix, H T is the transposed matrix of H, z k is the target observation result.

8. The real-time target detection and tracking method based on laser point cloud according to claim 6, characterized in that: The method of estimating the target trajectory by using the Kalman filter algorithm further includes: Update the error covariance matrix P at the current moment k k , and its calculation formula is 9. A real-time target detection and tracking system based on laser point cloud, characterized in that: include: A point cloud data acquisition module is used to acquire initial point cloud data through digital signals received by the laser radar; A preprocessing module preprocesses the initial point cloud data, including: performing voxel filtering on the initial point cloud data, filtering the initial point cloud data using the voxel filtering; performing ground filtering on the point cloud data after the voxel filtering to obtain first point cloud data; wherein the ground filtering includes: dividing the point cloud data after the voxel filtering into grids on the xy plane, and independently performing height estimation on each grid, the height estimation including maximum height and minimum height; assuming that the ideal ground height is h0, the grids that meet the following two conditions are removed as ground: the maximum height and minimum height of the grid are both within the range of [h0-Δ1,h0+Δ1]; the difference between the maximum height and the minimum height of the grid is less than Δ2, wherein h0 is the ideal ground height, Δ1 is the height difference threshold, and Δ2 is the maximum height and minimum height difference threshold of the grid; A clustering module is used to apply fast Euclidean clustering to the ground filtered point cloud data to obtain clustered point cloud data; the fast Euclidean clustering includes the following steps: constructing a balanced binary tree for the first point cloud data to accelerate the search; creating a labels data group and a removed data group that are the same size as the first point cloud data; creating an empty adjacency list G; traversing all points in the first point cloud data, merging the label numbers of the points according to the adjacency list G, assigning the same label number information to the points with connected label numbers in the adjacency list G, and outputting the label number array of the first point cloud data; obtaining the clustered point cloud data according to the label number array; wherein, traversing all points in the point cloud data includes: if the value of the removed data group corresponding to the current point p is true, the current point p has been processed and is skipped; creating a new label label p , label p Add to the adjacency list G; take the current point p as the center and r as the radius to perform a radius search to obtain a list of points adjacent to p; loop through the list of adjacent points to extract point q: when the label number of point q is label q Valid and not equal to the label number label of the point p p When label is created in the adjacency table G q and label p connection relationship; set the label number of the point q to label p , and set the removed value of the point q to true; A detection target list acquisition module, configured to acquire a detection target list based on the clustered point cloud data; A matching module, which associates and matches the existing trajectory with the detection target; The target trajectory acquisition module uses the Kalman filter algorithm to estimate the target trajectory.

10. An electronic device, characterized in that: include: Memory for storing computer programs; A processor, configured to execute the steps of the method according to any one of claims 1 to 8 when executing the computer program stored in the memory.

11. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are performed.

Images