All-terrain maneuvering target tracking system and method

Abstract

The invention discloses an all-terrain maneuvering target tracking system and an all-terrain maneuvering target tracking method, belongs to the technical field of all-terrain moving target tracking, and aims to solve the problems of poor target tracking trajectory adaptability and large vibration interference in a complex terrain. During operation, the target orientation is positioned, the advancing direction is updated in real time, after the collection range is dynamically delimited, multiple frames of images are collected, and a topographic map is constructed through cutting, splicing and distortion correction; preprocessing the topographic map, extracting landform features, and dividing passable areas; generating a tracking trajectory in the passable area, and optimizing the trajectory through the position change of the target; and adaptively adjusting parameters of the balance component according to the track area, and fusing a vibration signal to judge whether to trigger track updating so as to ensure that the track dynamically adapts to terrain and target movement. According to the invention, through enhancing image processing and multi-source data fusion, the continuity, accuracy and environmental adaptability of target tracking in a complex terrain are significantly improved, and the reliability and efficiency of operation are effectively enhanced.

Description

Technical Field

[0001] This invention relates to the field of all-terrain mobile target tracking technology, and more specifically to an all-terrain mobile target tracking system and method. Background Technology

[0002] When conducting target tracking operations in all-terrain environments, the complex and ever-changing terrain conditions pose numerous challenges to the stable operation of the tracking system. In existing technologies, target tracking trajectory planning often fails to adequately adapt to real-time terrain changes. When the terrain experiences sudden changes in slope or uneven obstacle distribution, the trajectory is prone to deviating from the actual passable path, leading to interruptions or decreased efficiency in the tracking process. Simultaneously, vibration interference caused by terrain undulations is significant. This vibration not only affects the vehicle's stability but may also reduce the accuracy of target detection equipment, resulting in errors in target orientation judgment. Furthermore, the target's orientation change during movement is uncertain, and existing tracking systems do not respond promptly enough to target movement, and their trajectory update mechanisms are not flexible enough, often resulting in a lack of synchronization between the trajectory and the target's actual movement path. These factors collectively make it difficult to effectively guarantee the continuity and accuracy of all-terrain target tracking operations, significantly limiting operational reliability and efficiency. Therefore, to overcome these limitations, this invention proposes an all-terrain mobile target tracking system and method. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide an all-terrain mobile target tracking system and method, solving the problem of how to improve the adaptability of target tracking trajectory to terrain and target movement in all-terrain environments, while reducing the interference of vibration on tracking operations. To achieve the above objective, the present invention provides the following technical solution:

[0004] A target tracking method for all-terrain maneuvering includes:

[0005] When conducting target tracking operations, the forward direction is updated based on the target orientation obtained by fusing lidar reflection signals and infrared detection data, and the acquisition range is dynamically defined to acquire and crop multiple frames of images within the acquisition range along the forward direction to construct a topographic map.

[0006] Multi-scale preprocessing of the topographic map and extraction of geomorphic features are performed. Core feature parameters are extracted to construct a comprehensive feature matrix. The geomorphic feature change trend is analyzed by combining the topographic information from the previous trajectory planning. Feature regions of the topographic map are divided and comprehensive geomorphic evaluation values ​​are calculated to screen candidate passable areas. Passable areas of the topographic map are constructed through verification and fusion.

[0007] Within the passable area, grid areas within the passable area are selected as core feasible sub-areas by setting a screening threshold range. Grid units are divided and comprehensive passage costs are calculated. An evaluation function for path search is constructed, a tracking trajectory for the current collection area is generated, and changes in target orientation are monitored. Transitional grid units are selected through a two-factor screening method to optimize and correct the tracking trajectory.

[0008] Based on the terrain features of the area where the tracking trajectory is located, the parameters of the balancing component are adaptively adjusted, vibration signals are extracted in real time, and a comprehensive analysis is performed on the execution progress of the tracking trajectory and the change in the target orientation to determine whether the trajectory update mechanism is triggered, thereby delineating a new collection range and updating the tracking trajectory.

[0009] Specifically, the steps for dynamically defining the data collection range include:

[0010] Configure the evaluation time range, obtain the driving speed and target movement speed within the evaluation time range from the current time, fuse the target radial velocity measured by lidar, match and align according to timestamps, and apply a moving average algorithm to generate the baseline driving speed and baseline target movement speed of the average motion state within the evaluation time range.

[0011] Using the end point of the current tracking trajectory as the reference point, and combining the reference driving speed and the reference target movement speed, the trajectory verification distance is obtained by calculating the product of the speed difference between the two and the preset response time, and superimposing the product of the reference driving speed and the preset response time as the next trajectory length. The lateral width is calculated by weighted calculation of the minimum turning radius parameter and the terrain openness quantification value obtained by the lidar scan.

[0012] The next trajectory length and horizontal width parameters are used to generate the boundary of the acquisition range. A Cartesian coordinate system is established with the reference point as the origin and the forward direction as the vertical axis. The coordinates of the rectangular boundary of the acquisition range are obtained through coordinate calculation.

[0013] Specifically, the steps for acquiring and cropping multiple frames of images within the acquisition range along the forward direction include:

[0014] Obtain the path parameters of the current tracking trajectory, preset the camera shooting angle at each coordinate position of the tracking trajectory, match the scanning angle of the LiDAR, send a start command to the camera group, call the preset basic shooting parameter combination, dynamically adjust the real-time shooting angle of the camera, and keep it spatially synchronized with the LiDAR scanning range.

[0015] The total number of image acquisition frames is set based on the total length of the tracking trajectory and the displacement velocity, while the image acquisition frame rate is set in combination with the displacement velocity.

[0016] It receives the real-time forward direction and acquisition range, and combines the path parameters of the current tracking trajectory to convert the acquisition range boundary parameters into an image pixel coordinate range that changes with the tracking trajectory coordinate position, according to the mapping rules between the tracking trajectory coordinate position and the image pixel.

[0017] Based on the converted image pixel coordinate range and the current tracking trajectory coordinate position, the original image captured by the camera is cropped frame by frame to obtain each cropped image.

[0018] Specifically, the steps for constructing a topographic map include:

[0019] For each cropped image, feature point matching is performed based on the driving coordinate position of the tracking trajectory. The three-dimensional terrain feature points extracted by the LiDAR are fused together to extract the terrain feature points related to the trajectory in each cropped image. The displacement of terrain feature points between different cropped images is calculated based on the change of driving coordinates of the tracking trajectory. Based on the displacement of terrain feature points, multiple cropped images are stitched together according to the trajectory coordinate information to construct a terrain map and perform distortion correction and color consistency adjustment.

[0020] Specifically, the steps for delineating the characteristic regions of a topographic map include:

[0021] The current topographic map is preprocessed to identify and extract its geomorphic features, which are then integrated to construct an initial set of geomorphic features. These geomorphic features include: hard obstacle features, terrain slope features, ground bearing capacity features, and terrain undulation features.

[0022] Based on the initial set of geomorphic features, core feature parameters are extracted to construct a comprehensive geomorphic feature matrix;

[0023] Retrieve the comprehensive geomorphic feature matrix and the boundary of the passable area from the topographic map collected during the last tracking trajectory planning, and spatially align the topographic map collected during the current tracking trajectory with the topographic map collected during the last tracking trajectory planning through coordinate matching;

[0024] The current integrated geomorphic feature matrix is ​​compared with the integrated geomorphic feature matrix of the previous tracking trajectory planning. The gradient of the change of each core feature parameter is fitted and the gradient threshold range is set. The current topographic map is divided into grid areas according to the preset grid.

[0025] By assigning weights to each core feature parameter and performing weighted summation, a comprehensive geomorphological evaluation value is calculated for each grid area. This value is then used to cluster the grids and divide the topographic map into feature regions using a clustering algorithm.

[0026] Specifically, the steps for constructing a topographic map of traversable areas include:

[0027] The boundary of the passable area planned in the previous tracking trajectory is used as a reference baseline. It is spatially correlated with the comprehensive geomorphological assessment value in each grid area and a safety threshold is matched. This is used to filter feature areas and mark candidate passable areas in combination with spatial topology analysis.

[0028] The candidate passable regions are checked for trend consistency. The sliding window algorithm is used to traverse the candidate passable regions. The gradient difference of the core feature parameters of the adjacent grid regions of the candidate passable regions is calculated and compared with the preset core feature parameter continuity threshold. The local abrupt grid regions are identified by the region labeling algorithm based on eight-neighbor search.

[0029] Locally mutated mesh regions are removed from candidate passable regions and then smoothed out to generate passable regions.

[0030] Specifically, the steps for generating the tracking trajectory of the current acquisition area include:

[0031] Extract the comprehensive geomorphological assessment value of all grid areas within the passable area, calculate the mean and standard deviation of the comprehensive geomorphological assessment value, and use them to set the upper and lower limits of the screening threshold range to screen the grid areas within the passable area as core feasible sub-regions;

[0032] Using the current tracking trajectory endpoint as the starting coordinate, and the target's location coordinates within the data collection area as the endpoint coordinate, grid cells are divided within the core feasible sub-region.

[0033] The corresponding geomorphic features of each grid cell are extracted and standardized. The standardized geomorphic features are accumulated by configuring dynamic weight coefficients to obtain the basic passage cost of each grid cell. The location influence weight is calculated by using a distance decay function. The basic passage cost and the location influence weight are proportionally superimposed and adjusted to obtain the comprehensive passage cost of the grid cell.

[0034] The comprehensive toll cost and the Euclidean distance between grid cells are weighted and integrated according to a set ratio to construct an evaluation function for path search, and a dynamically weighted path search is performed to generate multiple candidate paths;

[0035] Calculate the total length of each candidate path and compare it with the trajectory length of the current collection area to filter and obtain the tracking trajectory of the current collection area.

[0036] Specifically, the steps for optimizing and correcting the tracking trajectory include:

[0037] The tracking trajectory is smoothed and optimized, and the target's position change is monitored in real time. When the target's position change exceeds a preset deviation threshold, the tracking trajectory correction mechanism is triggered.

[0038] The coordinates and direction vector of the current tracking trajectory endpoint are retrieved, and the turning angle is calculated in combination with the target orientation. Transitional grid cells are selected in the core feasible sub-region using a two-factor screening method based on the turning angle and comprehensive travel cost. Transitional tracking trajectory segments are fitted by polynomial interpolation to optimize the tracking trajectory in the current acquisition area.

[0039] Specifically, the steps to determine whether the trajectory update mechanism has been triggered include:

[0040] A linkage judgment index system is set up. The linkage judgment index system is a multi-dimensional evaluation framework that integrates the target orientation deviation threshold, vibration signal change threshold, and trajectory execution progress threshold. It is used to comprehensively determine whether the current tracking trajectory is suitable for the target movement state and terrain features. It is constructed by associating the target movement orientation, terrain features, and tracking trajectory execution degree.

[0041] The target orientation deviation threshold is used to define the critical value at which the target has exceeded the current tracking trajectory's field of view and the trajectory needs to be forcibly updated;

[0042] The vibration signal change threshold is used to reflect the degree of mismatch between the current tracking trajectory and the actual terrain features;

[0043] The trajectory execution progress threshold is used to determine the lifecycle stage of the currently tracked trajectory;

[0044] Collect target location and vibration data, and record the distance ratio between the already tracked trajectory and the currently planned tracking trajectory. Through a linkage judgment indicator system, determine the trigger for the trajectory update mechanism.

[0045] A target tracking system for all-terrain mobility includes: a azimuth update module, a range delineation module, a terrain construction module, a region division module, a trajectory optimization module, a balance adjustment module, and an update judgment module;

[0046] The orientation update module tracks the target's orientation and adjusts the direction of travel in real time; the range delineation module dynamically delineates the acquisition range; the terrain construction module controls the camera to acquire and process images to construct a topographic map; the region division module preprocesses the topographic map, extracts landform features, constructs a comprehensive feature matrix, and divides traversable areas; the trajectory optimization module generates and optimizes the tracking trajectory in traversable areas; the balance adjustment module adaptively adjusts the balance component parameters based on the landform features of the tracking trajectory area; and the update judgment module comprehensively considers vibration signals, trajectory execution progress, and changes in target orientation to determine whether to trigger a trajectory update.

[0047] The beneficial effects of this invention are:

[0048] This application effectively improves the adaptability of target tracking trajectories in all-terrain environments by dynamically adapting to terrain and target movement status. It solves the problem of trajectory deviation from the actual passable path in existing technologies, ensuring the continuity of the tracking process, reducing interruptions, and improving operational efficiency. At the same time, by adaptively adjusting the parameters of the balancing component, it reduces vibration interference caused by terrain undulations, which helps maintain the stability of the vehicle and the accuracy of the target detection equipment, reducing the error in target orientation judgment. In addition, the flexible trajectory update mechanism can respond to changes in target movement status in a timely manner, avoiding the problem of trajectory being out of sync with the actual target movement path, thus enhancing the accuracy, reliability, and efficiency of all-terrain target tracking operations. Attached Figure Description

[0049] Figure 1 This is a flowchart of a target tracking method for all-terrain maneuvering according to the present invention;

[0050] Figure 2 A flowchart illustrating the specific steps involved in constructing a topographic map according to this invention;

[0051] Figure 3 This is a flowchart illustrating the specific steps involved in generating the tracking trajectory within the current acquisition range according to the present invention;

[0052] Figure 4 This is a schematic diagram of an all-terrain maneuverable target tracking system according to the present invention. Detailed Implementation

[0053] Please see Figure 1 This embodiment introduces a target tracking method for all-terrain maneuvering, including:

[0054] Step S1: During target tracking, the forward direction is updated in real time based on the target's location. The acquisition range is dynamically defined based on the current driving speed and the target's movement speed. This is used to control the camera to capture and crop multiple frames of images within the acquisition range along the forward direction to construct a terrain map. Specifically, a laser radar emits a laser beam and receives reflected signals. Combined with an infrared detector, the real-time orientation parameters of the target are continuously acquired. Combined with the vehicle's own GPS positioning information, a triangulation algorithm is used to calculate the relative angle between the target and the vehicle, which serves as the reference for the forward direction. When the target's orientation changes in real time, this relative angle is updated synchronously to adjust the forward direction. Simultaneously, statistical data on the vehicle's driving speed and the target's movement speed within the current time period are retrieved. The laser radar echo signal intensity distribution characteristics are fused, and a moving average algorithm is used to generate an overall speed index for both. Based on this index, the acquisition range is defined. Starting from the current tracking trajectory endpoint, a set trajectory length is extended in the forward direction as the longitudinal length. The lateral width is dynamically adapted based on the vehicle's turning radius and the terrain openness data acquired by the laser radar scan, ensuring that the acquisition range always covers the area required for the trajectory extension. The high-definition camera group at the front of the vehicle is activated. The camera group captures images of the foreground direction with a wide field of view. During the shooting process, the parameters are kept fixed, and the lidar point cloud data corresponding to each frame of the image is recorded synchronously. The acquired images are first cropped according to the real-time forward direction and the defined acquisition range. Then, feature points are extracted by matching the features of the lidar point cloud to achieve real-time stitching. After distortion correction by a preset lens distortion model, a complete topographic map is formed. The terrain roughness information retrieved by the lidar echo is superimposed and stored in the data processing unit for subsequent steps.

[0055] In this embodiment, by leveraging the wave reflection characteristics of lidar and the collaborative work of infrared detectors, combined with a triangulation algorithm, the system can accurately capture real-time changes in the target's location and synchronously adjust the direction of travel, ensuring that the vehicle always moves towards the target's location. The acquisition range is defined based on an overall velocity index generated by a moving average algorithm, incorporating terrain spatial features from laser detection. This avoids interference from real-time velocity fluctuations in the range definition, ensuring the acquisition range stably covers the area required for trajectory extension and providing sufficient terrain information for subsequent trajectory updates. A high-definition camera group captures images with a wide field of view and fixed parameters. Through the fusion of laser point clouds and images, and further processing such as cropping, stitching, and distortion correction, the system ensures comprehensive image acquisition and that the generated terrain map accurately reflects the terrain conditions within the acquisition range. This enhances the image's ability to monitor terrain details. After storage, it can directly provide high-quality data support for subsequent terrain feature extraction and trajectory planning, effectively ensuring the continuity and timeliness of the tracking process.

[0056] Preferably, the specific steps for dynamically defining the data collection range include:

[0057] The evaluation time range is configured based on the motion characteristics of the vehicle and the target to cover the time period that reflects the recent motion state of both, ensuring that the acquired speed data is representative. The vehicle speed and target movement speed within the evaluation time range from the current time are acquired and matched with the target radial speed measured by the lidar according to the timestamp. This ensures that each time point corresponds to a set of vehicle and target speed data, providing a unified time reference data source for subsequent calculations.

[0058] After the time axis is aligned, the driving speed and the target moving speed are processed by the moving average algorithm to eliminate the influence of instantaneous speed fluctuations. By calculating the average value of the driving speed and the target moving speed point by point within the evaluation time range, a benchmark driving speed and a benchmark target moving speed that can represent the average motion state of the vehicle and the target within the evaluation time range are generated as the core components of the overall speed index.

[0059] Using the endpoint of the current tracking trajectory as a reference point, and combining the reference driving speed and the reference target moving speed, the trajectory verification distance is obtained by calculating the product of the speed difference between the two and the preset response time, and superimposing the product of the reference driving speed and the preset response time. This is used to determine the longitudinal length of the acquisition range extending in the forward direction, ensuring that this length can cover the trajectory extension area required for the vehicle to track the target within the preset response time.

[0060] The minimum turning radius parameter is obtained through the vehicle's steering mechanism. The lateral passable width of the terrain within the acquisition range is measured and quantified using LiDAR scanning data to obtain a quantified value of terrain openness. The lateral width is calculated by weighting the minimum turning radius parameter and the quantified value of terrain openness. When the quantified value of terrain openness is higher than a preset upper threshold, the lateral width is increased proportionally to the difference between the quantified value and the upper threshold; when the quantified value of terrain openness is lower than a lower threshold, the lateral width is decreased proportionally to the difference between the lower threshold and the quantified value. The minimum turning radius parameter refers to the radius of the outer steering wheel's center trajectory when the vehicle turns at its maximum turning angle. It serves as the basic reference value for calculating the lateral width, ensuring that the vehicle does not exceed the acquisition range when turning. The upper and lower thresholds of openness are set based on the vehicle's maximum passable width and the lateral space required for normal turning.

[0061] The next trajectory length and horizontal width parameters are used to generate the boundary of the acquisition range. A Cartesian coordinate system is established with the reference point as the origin and the forward direction as the vertical axis. The coordinates of the four vertices of the rectangular boundary of the acquisition range are obtained through coordinate operations. The boundary is verified by combining the image pixel coordinate mapping relationship. The boundary coordinates are then checked and verified.

[0062] Please see Figure 2 Preferably, the specific steps for constructing a topographic map include:

[0063] The path parameters of the current tracking trajectory are obtained, including the trajectory direction, curvature change and the endpoint of the tracking trajectory. Based on the path parameters, the camera shooting angle at each coordinate position of the tracking trajectory is preset and matched with the scanning angle of the LiDAR. This ensures that the spatial mapping relationship between the initial shooting range and the acquisition range is compatible, providing basic parameters for trajectory adaptation for subsequent image acquisition.

[0064] The system sends a start command to the camera group, calls the preset basic shooting parameter combination, including focal length, exposure time, white balance and field of view, and receives the real-time coordinate position of the tracking trajectory to obtain the camera shooting angle. It dynamically adjusts the real-time shooting angle of the camera to ensure that the camera always faces the direction of the trajectory extension and maintains spatial synchronization with the LiDAR scanning range, thus ensuring dynamic matching between the shooting range and the acquisition range.

[0065] The total number of image acquisition frames is set according to the total length of the tracking trajectory and the displacement speed. The total number of frames is determined by the ratio of the total trajectory length to the trajectory length covered by a single frame image. At the same time, the image acquisition frame rate is set in combination with the displacement speed to ensure that each frame image corresponds to a complete set of laser echo data. This ensures that the images acquired when the vehicle moves forward have a consistent imaging benchmark, and that the frame rate and frame number are matched to avoid image overlap, redundancy, or information loss.

[0066] It receives the real-time forward direction and the generated acquisition range, and combines the path parameters of the current tracking trajectory. It then converts the boundary parameters of the acquisition range into an image pixel coordinate range that varies with the tracking trajectory coordinate position according to the mapping rules between the tracking trajectory coordinate position and the image pixel, thus providing a precise range for image cropping at different trajectory positions.

[0067] Based on the converted image pixel coordinate range and the current tracking trajectory coordinate position, the original image captured by the camera is cropped frame by frame to obtain each cropped image. The image portion within the acquisition range and corresponding to the trajectory position is retained, and the trajectory coordinate information and laser point cloud feature parameters corresponding to each cropped image are recorded simultaneously.

[0068] For each cropped image, feature point matching is performed based on the driving coordinate position of the tracking trajectory. The three-dimensional terrain feature points extracted by the LiDAR are fused together to extract the terrain feature points related to the trajectory in each cropped image. The displacement of terrain feature points between different cropped images is calculated based on the change of driving coordinates of the tracking trajectory. Based on the displacement of terrain feature points, multiple cropped images are stitched together according to the trajectory coordinate information to construct a terrain map that covers the acquisition range and matches the trajectory.

[0069] The system calls a preset lens distortion correction model, which contains distortion parameters corresponding to the camera's shooting angle. The corresponding parameters are selected based on the real-time shooting angle to correct the distortion of the stitched terrain map and eliminate image distortion caused by angle changes.

[0070] The distortion-corrected images undergo color consistency adjustment to unify the tones of images acquired at different trajectory locations. The continuity of image stitching is then verified by trajectory coordinates, and the consistency of terrain details is verified by combining the laser echo signal intensity distribution. Discontinuous areas of stitching are smoothed out, ultimately forming a complete topographic map, which is stored in the data processing section and associated with the corresponding tracking trajectory parameters for subsequent steps.

[0071] Step S2: Preprocess the current topographic map and extract geomorphic features. Extract core feature parameters to construct a comprehensive feature matrix. Combine the topographic information from the previous trajectory planning to analyze the changing trends of geomorphic features. Divide the topographic map into feature regions and calculate a comprehensive geomorphic evaluation value to screen candidate passable areas. Verify and fuse the data to construct passable areas on the topographic map and associate them with tracking trajectory information. Specifically, after preprocessing the current topographic map, fuse LiDAR point cloud data and extract various geomorphic features, including hard obstacle features, terrain slope features, ground bearing capacity features, and terrain undulation features. Extract core feature parameters to construct a comprehensive feature matrix to quantify terrain features. Combine the topographic information from the previous trajectory planning, establish temporal correlation through spatial alignment, and analyze the changing trends of geomorphic features by combining the terrain detail changes detected in the image. Divide the topographic map into feature regions based on the changing trends and calculate a comprehensive geomorphic evaluation value. Refer to the boundaries of the passable areas from the previous time and combine them with the vehicle's passability to screen candidate passable areas. The candidate region is checked for trend consistency. The verification is assisted by the distribution of laser echo signal intensity. After removing the parts that do not conform to the trend, the region is merged with the previously passable region to form the current passable region. The output dataset containing relevant parameters is linked with the tracking trajectory information for use in subsequent steps.

[0072] In this embodiment, preprocessing the current topographic map and extracting various geomorphic features, including hard obstacles and terrain slope, using laser detection data provides comprehensive and accurate basic data for subsequent analysis. Constructing a comprehensive feature matrix by selecting core feature parameters enables quantitative representation of terrain features, significantly improving the efficiency and accuracy of data processing. Combining the terrain information from the previous trajectory planning, spatial alignment is used to establish temporal correlations to analyze the changing trends of geomorphic features, enhancing the dynamic capture capability of image monitoring. This fully utilizes the extensibility of terrain changes, making terrain judgments more consistent with reality and reducing biases caused by isolated analysis of the current terrain. Dividing the topographic map into feature regions based on changing trends and calculating comprehensive geomorphic evaluation values, while referencing historical passable area boundaries and combining vehicle passability to screen candidate passable areas, effectively narrows the scope and improves the targeting and reliability of passable area selection. The candidate region is checked for trend consistency and merged with historical passable regions to form the current passable region. The continuity and safety of the passable region are further ensured by dual verification of image and laser data. The output dataset of associated tracking trajectory information provides high-quality data support for subsequent trajectory planning, which significantly improves the efficiency and accuracy of trajectory planning in target tracking operations.

[0073] Preferably, the specific steps for delineating the passable area include:

[0074] The current topographic map is preprocessed, including noise filtering and contrast enhancement. The image grayscale value is corrected by combining the intensity of the LiDAR reflection signal. A multi-scale feature extraction algorithm is used to scan the current topographic map, identify and extract the geomorphic features of the current topographic map, including hard obstacle features, terrain slope features, ground bearing capacity features and terrain undulation features. The three-dimensional coordinate information of the LiDAR point cloud is simultaneously associated and integrated to construct an initial geomorphic feature set.

[0075] Based on the initial set of geomorphic features, principal component analysis is used to extract core feature parameters that significantly affect passage, including the spatial proportion of obstacles, the mean and variance of slope, the uniformity of texture and the frequency of undulation. These parameters are then incorporated into the terrain roughness parameters obtained from laser echo inversion to construct a comprehensive geomorphic feature matrix, thereby achieving a quantitative representation of the geomorphic features of the topographic map.

[0076] The topographic information of the topographic map collected in the previous tracking trajectory planning is retrieved, including the comprehensive geomorphic feature matrix and the boundary of the passable area. The topographic map collected in the current tracking trajectory is spatially aligned with the topographic map collected in the previous tracking trajectory planning through coordinate matching. The laser point cloud data collected in the two times are superimposed for deviation analysis, and the temporal correlation of the topographic information is established to provide basic data for the analysis of geomorphic feature change trends.

[0077] The difference between the current comprehensive geomorphic feature matrix and the comprehensive geomorphic feature matrix of the previous tracking trajectory planning is calculated. Combined with the terrain extension law and the dynamic changes of terrain features in the image sequence, the gradient of the change of each core feature parameter is obtained by fitting the trend line, so as to determine the trend direction and strength of terrain change.

[0078] Based on the gradient changes of each core feature parameter, the current topographic map is divided into grid regions according to a preset grid by setting a gradient threshold range. By assigning weights to each core feature parameter and performing weighted summation, the comprehensive geomorphological evaluation value of each grid region is calculated. The grid regions with similar comprehensive geomorphological evaluation values ​​and spatial continuity are grouped into the same feature region by a clustering algorithm. Each feature region is assigned a label containing the dominant terrain type and change trend.

[0079] The boundary of the passable area planned in the previous tracking trajectory is used as a reference baseline. The comprehensive geomorphological evaluation value of each grid area is spatially correlated with the reference baseline, and a safety threshold is matched according to the vehicle's passability parameters. Feature areas are screened based on the safety threshold. The connectivity between the feature areas and the reference baseline is checked through image pixel-level connectivity analysis and spatial topology analysis. Feature areas whose evaluation values ​​meet the safety threshold and have a connecting path with the reference baseline are marked as candidate passable areas.

[0080] The candidate passable areas are checked for trend consistency. A sliding window algorithm is used to traverse the candidate passable areas. The gradient difference of the core feature parameters of the adjacent grid areas of the candidate passable areas is calculated and compared with the preset core feature parameter continuity threshold to check the curve continuity. Combined with the real-time terrain data of LiDAR scanning, a region labeling algorithm based on eight neighborhood search is used to identify local abrupt grid areas. That is, starting from the abnormal grid area where the gradient difference of the core feature parameter is greater than the core feature parameter continuity threshold, the abnormal grid areas in the surrounding eight neighborhoods are traversed. The continuous abnormal grid areas are marked as the same local abrupt area. Then, the elimination mechanism is triggered to remove the local abrupt grid area from the candidate passable area.

[0081] The boundaries of candidate passable regions with removed local mutation areas are smoothed to generate passable regions. The boundary accuracy is optimized by image edge detection. The boundary coordinates of passable regions, comprehensive landform features and change trend parameters are integrated into a dataset and bound to the corresponding tracking trajectory information fields.

[0082] Step S3: Within the passable area, the grid areas within the passable area are selected as core feasible sub-areas by setting a screening threshold range. The grid cells are divided and the comprehensive passage cost is calculated. An evaluation function for path search is constructed. The path planning algorithm is called to generate the tracking trajectory of the current collection area. The target orientation change is monitored by combining the real-time echo data of the lidar. Transition grid cells are selected by the two-factor screening method to optimize the tracking trajectory. Specifically, using the coordinates corresponding to the current tracking trajectory endpoint and the real-time azimuth of the target within the acquisition area as the starting and ending points, grid cells are divided within the core feasible sub-region. The grid cell size is set according to the trajectory accuracy requirements. The terrain features associated with each grid cell are extracted and standardized. The basic passage cost is calculated by fusing terrain hardness parameters retrieved from laser point clouds and then incorporating relative position factors to obtain the comprehensive passage cost. An evaluation function is constructed to call a path planning algorithm to search for paths. The tracking trajectory in the current acquisition area is selected and smoothed by combining the continuity analysis of terrain features in the image sequence. The target azimuth changes are monitored in real time. When the change exceeds a preset deviation threshold, a correction mechanism is triggered. A transition grid cell is selected using a two-factor screening method to fit the transition trajectory segment and optimize the current tracking trajectory to achieve smooth connection. At the same time, the grid cell size is dynamically adjusted according to the trajectory length and the vehicle's computing power. The optimized tracking trajectory is converted into a recognizable coordinate sequence. The image pixel coordinate mapping relationship is superimposed, verified, and stored in the trajectory database with corresponding information for retrieval.

[0083] In this embodiment, by setting a filtering threshold range to select core feasible sub-regions, it is possible to accurately focus on areas with better terrain conditions, reduce the interference of unnecessary terrain data on trajectory planning, and lay a high-quality data foundation for subsequent path search. Dividing the grid cells according to accuracy requirements balances the accuracy and computational efficiency of trajectory planning. The calculation of comprehensive travel costs integrates geomorphic features and location factors, incorporating terrain physical attribute parameters from laser detection to make cost assessment more aligned with actual travel needs, thus improving the rationality of path planning. The path planning algorithm, combined with smoothing optimization processing, ensures that the generated tracking trajectory is both terrain-adaptable and... The system reduces vehicle steering adjustments, ensuring smooth driving. A target orientation change monitoring and correction mechanism, based on the fusion positioning of LiDAR and image data, uses a two-factor selection process to fit transitional trajectory segments with transitional grid cells, achieving dynamic trajectory adaptation and ensuring continuous and accurate tracking even when the target is moving. Dynamic adjustment of grid cell size balances computational load while maintaining trajectory accuracy. A standardized data storage and association mechanism, binding LiDAR echo features and image frame indexes, provides reliable support for real-time trajectory retrieval and subsequent analysis, comprehensively improving the efficiency, accuracy, and adaptability of tracking trajectory planning.

[0084] Please see Figure 3 Preferably, the specific steps for generating the tracking trajectory within the current acquisition range include:

[0085] The comprehensive geomorphological assessment values ​​of all grid areas within the passable area are extracted. Combined with the terrain roughness distribution obtained from LiDAR scanning, the mean and standard deviation of these comprehensive geomorphological assessment values ​​are calculated. These values ​​are used to set the upper and lower limits of the screening threshold range, thus selecting grid areas within the passable area as core feasible sub-regions. Specifically, the lower limit of the basic screening threshold is set by subtracting a certain number of standard deviations from the mean, and the upper limit is set by adding a certain number of standard deviations to the mean, forming the initial threshold range. Considering the total area of ​​the passable area and the number of grid units, when the total number of grid areas exceeds a preset threshold, the lower limit of the basic screening threshold is increased to reduce the number of areas to be screened; when the total number of grid areas is lower than the preset threshold, the upper limit of the basic screening threshold is decreased to increase the number of areas to be screened, dynamically adapting to the current scale of the passable area.

[0086] Using the current tracking trajectory endpoint as the starting coordinate, the coordinates corresponding to the real-time orientation of the target within the acquisition area are used as the endpoint coordinates. The core feasible sub-region is divided into grid cells, and the grid cell size is set according to the trajectory accuracy requirements. The higher the accuracy requirement, the smaller the grid cell size. The corresponding image pixel block features are synchronously associated.

[0087] The corresponding geomorphic features of each grid cell are extracted and standardized. The standardized geomorphic features are then accumulated using dynamic weighting coefficients to obtain the basic travel cost for each grid cell. A distance decay function is used to calculate the location influence weight. The basic travel cost and the location influence weight are then proportionally combined and adjusted to obtain the comprehensive travel cost of the grid cell. Grid cells closer to the starting or ending point have lower location influence weights to prioritize the smooth connection between the starting and ending points of the path.

[0088] The overall travel cost and the Euclidean distance between grid cells are weighted and fused according to a set ratio to construct an evaluation function for path search. Based on this function, a path planning algorithm is invoked to perform a dynamically weighted path search. During the search process, the path planning algorithm prioritizes the grid cell with the lowest evaluation function value as the path node. After generating multiple candidate paths, the feasibility of the paths is verified by predicting the motion trajectory of the image sequence. The total length of each candidate path is calculated and compared with the trajectory length of the current acquisition area. The candidate path with the matching length and the fewest corners is selected as the tracking trajectory for the current acquisition area.

[0089] The tracking trajectory is smoothed and optimized by fitting a cubic Bézier curve to ensure that the rate of change of the curve curvature does not exceed the maximum tolerance of the vehicle steering system, eliminating sharp corners in the trajectory, and verifying the fit between the curve and the terrain by using laser point cloud density.

[0090] The system monitors target azimuth changes in real time. Based on the fusion positioning results of LiDAR and imagery, a tracking trajectory correction mechanism is triggered when the target azimuth change exceeds a preset deviation threshold. It retrieves the coordinates and direction vector of the current tracking trajectory endpoint, calculates the turning angle based on the target azimuth, and selects transitional grid cells within the core feasible sub-region using a two-factor screening method based on the turning angle and overall travel cost. Specifically, it first filters grid cells within the direction range matching the turning angle, then selects the grid cell with the lowest overall travel cost as the transition node. Referring to the terrain continuity features in the image, a transitional tracking trajectory segment is fitted using polynomial interpolation to optimize the tracking trajectory in the current acquisition area, ensuring that the tangent direction of the transitional trajectory segment is consistent with the current tracking trajectory endpoint for a smooth transition.

[0091] The grid cell size is dynamically adjusted based on the trajectory length of the current acquisition area and the vehicle's computing power. When the trajectory length is long, the grid cell size is increased to reduce computational load; when the trajectory is close to the target and accuracy requirements increase, the grid cell size is reduced to improve trajectory detail. The adjusted grid cells need to be re-associated with terrain feature parameters and the overall travel cost is calculated. The terrain adaptability parameters are updated in conjunction with the laser echo signal intensity to ensure adaptability to terrain features.

[0092] The optimized tracking trajectory of the current acquisition area is converted into a recognizable coordinate sequence, including the coordinates of the tracking trajectory position, the turning angle, and the suggested driving speed. The continuity and feasibility of the trajectory are checked through a dual verification mechanism of image and laser data. After confirmation, the trajectory is stored in the trajectory database and associated with the corresponding target orientation information and terrain parameters for real-time retrieval.

[0093] Step S4: When the vehicle travels along the planned tracking trajectory, the balance component parameters are adaptively adjusted based on the terrain features of the area where the tracking trajectory is located to suppress the transmission of road vibrations caused by terrain changes. Simultaneously, vibration signals are extracted in real time, and a comprehensive analysis is performed combining the changes in lidar echo signal intensity and terrain feature offsets in the image frame sequence, along with the tracking trajectory execution progress and target orientation changes, to determine whether to trigger the trajectory update mechanism. A new acquisition range is then defined, and a corresponding tracking trajectory is generated, ensuring that the trajectory adapts synchronously to target movement and terrain changes. Specifically, vibration data is acquired and quantified through vibration sensors, and the balance component is dynamically adjusted based on the current road section's terrain features to maintain vibration attenuation. The trajectory execution progress and target orientation changes are monitored in real time, and the actual terrain changes are calculated through image feature matching. Multi-level thresholds are set to determine whether to trigger an update, and an early warning or immediate update is initiated if necessary. After an update is triggered, the forward direction is adjusted based on the current position, and a new acquisition range is defined based on speed indicators and terrain conditions. Simultaneously, the camera group and lidar are activated for collaborative acquisition, collecting terrain data to generate a terrain map. Based on the previous logic, terrain features are extracted, passable areas are delineated, and a new tracking trajectory is planned.

[0094] In this embodiment, the vehicle adaptively adjusts the parameters of the balance component based on the terrain features of the trajectory area during operation. This effectively suppresses road vibration transmission caused by terrain changes, reducing the impact of vibration on vehicle stability and equipment accuracy, and improving the smoothness of the driving process. Real-time extraction of vibration signals, combined with trajectory execution progress and target orientation changes, triggers a multi-dimensional trajectory update mechanism. This avoids the limitations of relying on a single indicator and accurately captures scenarios requiring trajectory updates, ensuring the rationality and timeliness of the update timing. The process of adjusting the direction of travel from the current position, redefining the acquisition range, and generating a new tracking trajectory achieves dynamic adaptation between the trajectory and the target's movement state and terrain changes. This ensures the vehicle always tracks the target on the optimal path, effectively improving the continuity, accuracy, and environmental adaptability of the tracking process, and overall enhancing the reliability and efficiency of target tracking operations.

[0095] Preferably, the specific steps for adaptively adjusting the balance component parameters based on the terrain features of the area where the tracking trajectory is located include:

[0096] Retrieve the comprehensive geomorphic feature matrix corresponding to the current tracking trajectory, extract the core geomorphic parameters of each grid area within the trajectory coverage range, including terrain slope, ground bearing capacity, undulation frequency and obstacle distribution density, establish a mapping relationship model between core geomorphic parameters and balance component parameters, and clarify the benchmark adjustment range corresponding to different parameter intervals.

[0097] The vehicle positioning module obtains the current driving position coordinates in real time, matches them to the corresponding grid area, retrieves the core terrain parameters of the area, generates the initial adjustment values ​​of the balance component parameters based on the mapping relationship model, and sends them to the balance component control center.

[0098] Preferably, the specific steps for determining whether the trajectory update mechanism has been triggered include:

[0099] By combining the dynamic changes in terrain and the movement characteristics of the target during target tracking operations, a linkage judgment index system is established, focusing on target orientation deviation, terrain adaptability, and trajectory lifecycle. The linkage judgment index system is a multi-dimensional evaluation framework integrating target orientation deviation threshold, vibration signal change threshold, and trajectory execution progress threshold. It is used to comprehensively determine whether the current tracking trajectory is suitable for the target's movement state and terrain features, constructed by associating target movement orientation, terrain features, and tracking trajectory execution degree. Specifically, the target orientation deviation threshold defines the critical value at which the target has exceeded the current tracking trajectory's field of view and requires a forced trajectory update. This is set by analyzing the maximum deviation of the target's movement trajectory and the vehicle's tracking field of view, combined with safety redundancy. The vibration signal change threshold reflects the degree of mismatch between the current tracking trajectory and the actual terrain features. It is calculated by collecting vibration benchmark data when the vehicle travels in different terrain feature areas, calculating the normal fluctuation range of the vibration signal, and setting the critical value for abnormal changes. The trajectory execution progress threshold determines the lifecycle stage of the current tracking trajectory for advance planning and connection. It is set by analyzing the matching relationship between the total trajectory length and the target's movement speed, and setting warning nodes and update nodes according to the trajectory execution ratio.

[0100] The target's location is collected in real time and compared with the location reference value during the tracking trajectory planning to calculate the location deviation. When the location deviation exceeds the target location deviation threshold, the trajectory update mechanism is triggered to avoid target loss.

[0101] Vibration data is collected by vibration sensors, and the vibration deviation between the data and the reference vibration data is calculated. If the vibration deviation exceeds the vibration signal change threshold, the trajectory update mechanism is triggered. The reference vibration data refers to the vibration data benchmark value collected by the vehicle during simulated driving or historical actual driving in each grid area corresponding to the current planned trajectory, based on preset terrain feature parameters. The initial reference data is generated by retrieving the comprehensive terrain feature matrix generated during the trajectory planning stage, combining it with the comprehensive terrain evaluation value of each grid area, matching the preset vibration feature mapping model, and then dynamically calibrating it by integrating the vibration data from the previous actual driving of the trajectory.

[0102] The distance ratio between the already tracked trajectory and the currently planned tracking trajectory is dynamically recorded. When the distance ratio exceeds the trajectory execution progress threshold, the trajectory update mechanism is triggered to ensure that the trajectory continuously covers the target's movement path.

[0103] Please see Figure 4 This embodiment introduces an all-terrain mobile target tracking system, including a azimuth update module, a range delineation module, a terrain construction module, a region division module, a trajectory optimization module, a balance adjustment module, and an update judgment module.

[0104] The orientation update module is used to track the target's orientation and adjust the direction of travel in real time. It uses a lidar to emit laser beams and receive reflected signals, and an infrared detector to obtain the target's real-time orientation parameters. Combined with the vehicle's GPS positioning, a triangulation algorithm is used to calculate the relative angle between the target and the vehicle as a reference for the direction of travel. When the target's orientation changes, the module integrates the target's motion trend analysis from the image frame sequence and updates the relative angle synchronously to adjust the direction, ensuring that the vehicle always faces the target.

[0105] The range delineation module is used to dynamically define the acquisition range. It retrieves the current speed statistics of the vehicle and the target, integrates the radial velocity of the target measured by the lidar, and generates an overall speed index using a moving average algorithm. This index is used to define the acquisition range. Starting from the current trajectory endpoint, the longitudinal length is extended in the forward direction by a set trajectory length. The lateral width is adapted based on the vehicle's turning radius and the terrain openness obtained by the lidar scan to ensure coverage of the area required for the trajectory extension. Boundary verification is performed by combining the image pixel coordinate mapping relationship.

[0106] The terrain construction module is used to control the cameras to acquire and process images to build a terrain map. It starts the front-end high-definition camera group, synchronously recording the LiDAR point cloud data corresponding to each frame of image. Foreground images are captured from a wide field of view while maintaining fixed parameters, ensuring spatial synchronization with the LiDAR scanning range. The acquired images are cropped, stitched together using feature point matching with the LiDAR point cloud features, and lens distortion corrected. Then, terrain roughness information retrieved from the LiDAR echo is superimposed to form a complete terrain map, which is stored in the data processing unit for subsequent use.

[0107] The region segmentation module is used to preprocess the topographic map and extract geomorphic features, construct a comprehensive feature matrix, and analyze the changing trends by combining the topographic information from the previous tracking trajectory planning. Based on this, the topographic map is divided into feature regions and a comprehensive geomorphic evaluation value is calculated. Referring to the boundaries of historical passable areas and combining vehicle passability, candidate areas are selected, and after verification and fusion, the current passable area is formed, outputting a dataset of associated trajectory information.

[0108] The trajectory optimization module is used to generate and optimize tracking trajectories in passable areas. It selects core feasible sub-regions, divides them into grid cells, calculates the comprehensive travel cost, constructs an evaluation function, and calls a path planning algorithm to generate the trajectory. It monitors changes in target orientation, and when the deviation threshold is exceeded, it uses a two-factor filter to fit transitional grid cells to a transition segment, while dynamically adjusting the grid size to optimize the trajectory.

[0109] The balance adjustment module is used to adaptively adjust the parameters of the balance components based on the geomorphological features of the tracking trajectory area to suppress road vibration transmission. It retrieves the comprehensive geomorphological feature matrix of the current tracking trajectory, extracts the core geomorphological parameters of each grid, and establishes a mapping model with the balance component parameters. The positioning module matches the current grid, generates initial adjustment values ​​based on the model, and sends them to the control center for dynamic calibration using real-time vibration data.

[0110] The update judgment module is used to determine whether to trigger a trajectory update by comprehensively considering vibration signals, trajectory execution progress, and changes in target orientation. A linkage indicator system is set up to collect target orientation deviation, collect vibration data to calculate deviation, and record the percentage of the traveled trajectory in real time. If any indicator exceeds the threshold, an update is triggered, and a new collection range is defined and a corresponding trajectory is generated.

[0111] Working principle and its effects:

[0112] This invention achieves accurate target tracking in all terrain environments. The core of this invention lies in the integration of lidar and image data to construct a dynamic adaptation mechanism: First, the forward direction is adjusted in real time based on the target's orientation and terrain features, and the acquisition range is defined to generate a high-precision terrain map; then, passable areas are divided through geomorphological feature analysis, and the tracking trajectory is planned and optimized in combination with comprehensive travel costs; at the same time, the balancing component is adaptively adjusted according to terrain characteristics, and the trajectory update is determined through multi-dimensional indicators, ultimately achieving synchronous adaptation between the trajectory and target movement and terrain changes.

[0113] In the target localization and terrain modeling stage, the wave reflection characteristics of lidar and infrared detection work together to obtain the target's orientation. Combined with GPS positioning, the triangulation algorithm accurately locks the direction of travel. The dynamically defined acquisition range is integrated into the terrain spatial features detected by lidar, effectively avoiding interference from speed fluctuations. High-definition cameras and lidar work together to acquire images. After cropping, stitching, and distortion correction, a terrain map with superimposed terrain roughness information is generated. This ensures the comprehensiveness of the images and enhances the ability to monitor terrain details by fusing lidar point clouds with images. This provides high-quality data support for subsequent analysis and significantly improves the orientation accuracy and terrain characterization precision in the initial stage of tracking.

[0114] In the trajectory planning and dynamic optimization stages, topographic features are extracted by integrating laser point cloud data during topographic map preprocessing. A comprehensive feature matrix is ​​constructed and its changing trends are analyzed in conjunction with historical information. Feature regions are divided using clustering algorithms, and evaluation values ​​are calculated. Passable areas are screened through double verification to ensure regional continuity and safety. Within the passable areas, a tracking trajectory is generated based on the comprehensive travel cost and path evaluation function. The trajectory is smoothly optimized by using a two-factor selection of transitional grid cells in conjunction with changes in target orientation. The dynamically adjusted grid size balances accuracy and computational load, enabling the trajectory to adapt to the terrain and respond promptly to target movement, significantly improving the rationality and dynamic adaptability of path planning.

[0115] Overall, this invention achieves dynamic adaptation across the entire process from target localization and terrain modeling to trajectory planning and updating. It not only improves the adaptability of target tracking trajectories in complex terrain and reduces vibration interference, but also ensures the synchronization of trajectory with target movement and terrain changes, effectively enhancing the continuity, accuracy, reliability, and efficiency of all-terrain target tracking operations.

[0116] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A target tracking method for all-terrain maneuvering, characterized in that, include: When conducting target tracking operations, the forward direction is updated based on the target orientation obtained by fusing lidar reflection signals and infrared detection data, and the acquisition range is dynamically defined to acquire and crop multiple frames of images within the acquisition range along the forward direction to construct a topographic map. The topographic map is preprocessed at multiple scales and geomorphic features are extracted. Core feature parameters are extracted to construct a comprehensive feature matrix. The geomorphic feature change trend is analyzed by combining the topographic information from the previous trajectory planning. Feature regions of the topographic map are divided and comprehensive geomorphic evaluation values ​​are calculated to screen candidate passable areas. Passable areas of the topographic map are constructed through verification and fusion. Within the passable area, grid areas within the passable area are selected as core feasible sub-regions by setting a screening threshold range. Grid cells are divided and comprehensive passage costs are calculated. An evaluation function for path search is constructed, a tracking trajectory for the current collection area is generated, and changes in target orientation are monitored. Transitional grid cells are selected using a two-factor screening method to optimize and correct the tracking trajectory. Based on the terrain features of the area where the tracking trajectory is located, the parameters of the balance component are adaptively adjusted, vibration signals are extracted in real time, and the execution progress of the tracking trajectory and the change in the target orientation are comprehensively analyzed to determine whether the trajectory update mechanism is triggered, thereby defining a new collection range and updating the tracking trajectory.

2. The all-terrain maneuverable target tracking method as described in claim 1, characterized in that, The specific steps for dynamically defining the acquisition range include: Configure the evaluation time range, obtain the driving speed and target movement speed within the evaluation time range from the current time, fuse the target radial velocity measured by lidar, match and align according to timestamps, and apply a moving average algorithm to generate the baseline driving speed and baseline target movement speed of the average motion state within the evaluation time range. Using the end point of the current tracking trajectory as the reference point, and combining the reference driving speed and the reference target movement speed, the trajectory verification distance is obtained by calculating the product of the speed difference between the two and the preset response time, and superimposing the product of the reference driving speed and the preset response time as the next trajectory length. The lateral width is calculated by weighted calculation of the minimum turning radius parameter and the terrain openness quantification value obtained by the lidar scan. The next trajectory length and horizontal width parameters are used to generate the boundary of the acquisition range. A Cartesian coordinate system is established with the reference point as the origin and the forward direction as the vertical axis. The coordinates of the rectangular boundary of the acquisition range are obtained through coordinate calculation.

3. The all-terrain maneuverable target tracking method as described in claim 1, characterized in that, The specific steps for acquiring and cropping multiple frames of images within the acquisition range along the forward direction include: Obtain the path parameters of the current tracking trajectory, preset the camera shooting angle at each coordinate position of the tracking trajectory, match the scanning angle of the LiDAR, send a start command to the camera group, call the preset basic shooting parameter combination, dynamically adjust the real-time shooting angle of the camera, and keep it spatially synchronized with the LiDAR scanning range. The total number of image acquisition frames is set based on the total length of the tracking trajectory and the displacement velocity, while the image acquisition frame rate is set in combination with the displacement velocity. It receives the real-time forward direction and acquisition range, and combines the path parameters of the current tracking trajectory to convert the acquisition range boundary parameters into an image pixel coordinate range that changes with the tracking trajectory coordinate position, according to the mapping rules between the tracking trajectory coordinate position and the image pixel. Based on the converted image pixel coordinate range and the current tracking trajectory coordinate position, the original image captured by the camera is cropped frame by frame to obtain each cropped image.

4. The all-terrain maneuverable target tracking method as described in claim 3, characterized in that, The specific steps for constructing the topographic map include: For each cropped image, feature point matching is performed based on the driving coordinate position of the tracking trajectory. The three-dimensional terrain feature points extracted by the LiDAR are fused together to extract the terrain feature points related to the trajectory in each cropped image. The displacement of terrain feature points between different cropped images is calculated based on the change of driving coordinates of the tracking trajectory. Based on the displacement of terrain feature points, multiple cropped images are stitched together according to the trajectory coordinate information to construct a terrain map and perform distortion correction and color consistency adjustment.

5. The all-terrain maneuverable target tracking method as described in claim 1, characterized in that, The specific steps for dividing the topographic map into characteristic regions include: The current topographic map is preprocessed to identify and extract its geomorphic features, and then integrated to construct an initial set of geomorphic features. These geomorphic features include: hard obstacle features, terrain slope features, ground bearing capacity features, and terrain undulation features. Based on the initial set of geomorphic features, core feature parameters are extracted to construct a comprehensive geomorphic feature matrix; Retrieve the comprehensive geomorphic feature matrix and the boundary of the passable area from the topographic map collected during the last tracking trajectory planning, and spatially align the topographic map collected during the current tracking trajectory with the topographic map collected during the last tracking trajectory planning through coordinate matching; The current integrated geomorphic feature matrix is ​​compared with the integrated geomorphic feature matrix of the previous tracking trajectory planning. The gradient of the change of each core feature parameter is fitted and the gradient threshold range is set. The current topographic map is divided into grid areas according to the preset grid. By assigning weights to each core feature parameter and performing weighted summation, a comprehensive geomorphological evaluation value is calculated for each grid area. This value is then used to cluster the grids and divide the topographic map into feature regions using a clustering algorithm.

6. The all-terrain maneuverable target tracking method as described in claim 5, characterized in that, The specific steps for constructing the traversable area of ​​the topographic map include: The boundary of the passable area planned in the previous tracking trajectory is used as a reference baseline. It is spatially correlated with the comprehensive geomorphological assessment value in each grid area and a safety threshold is matched. This is used to filter feature areas and mark candidate passable areas in combination with spatial topology analysis. The candidate passable regions are checked for trend consistency. The sliding window algorithm is used to traverse the candidate passable regions. The gradient difference of the core feature parameters of the adjacent grid regions of the candidate passable regions is calculated and compared with the preset core feature parameter continuity threshold. The local abrupt grid regions are identified by the region labeling algorithm based on eight-neighbor search. Locally mutated mesh regions are removed from candidate passable regions and then smoothed out to generate passable regions.

7. The all-terrain maneuverable target tracking method as described in claim 1, characterized in that, The specific steps for generating the tracking trajectory of the current acquisition area include: Extract the comprehensive geomorphological assessment value of all grid areas within the passable area, calculate the mean and standard deviation of the comprehensive geomorphological assessment value, and use them to set the upper and lower limits of the screening threshold range to screen the grid areas within the passable area as core feasible sub-regions; Using the current tracking trajectory endpoint as the starting coordinate, and the target's location coordinates within the data collection area as the endpoint coordinate, grid cells are divided within the core feasible sub-region. The corresponding geomorphic features of each grid cell are extracted and standardized. The standardized geomorphic features are accumulated by configuring dynamic weight coefficients to obtain the basic passage cost of each grid cell. The location influence weight is calculated by using a distance decay function. The basic passage cost and the location influence weight are proportionally superimposed and adjusted to obtain the comprehensive passage cost of the grid cell. The comprehensive toll cost and the Euclidean distance between grid cells are weighted and integrated according to a set ratio to construct an evaluation function for path search, and a dynamically weighted path search is performed to generate multiple candidate paths; Calculate the total length of each candidate path and compare it with the trajectory length of the current collection area to filter and obtain the tracking trajectory of the current collection area.

8. The all-terrain maneuverable target tracking method as described in claim 7, characterized in that, The specific steps for optimizing and correcting the tracking trajectory include: The tracking trajectory is smoothed and optimized, and the target's position change is monitored in real time. When the target's position change exceeds a preset deviation threshold, the tracking trajectory correction mechanism is triggered. The coordinates and direction vector of the current tracking trajectory endpoint are retrieved, and the turning angle is calculated in combination with the target orientation. Transitional grid cells are selected in the core feasible sub-region using a two-factor screening method based on the turning angle and comprehensive travel cost. Transitional tracking trajectory segments are fitted by polynomial interpolation to optimize the tracking trajectory in the current acquisition area.

9. The all-terrain maneuverable target tracking method as described in claim 1, characterized in that, The specific steps for determining whether the trajectory update mechanism has been triggered include: A linkage judgment index system is set up. The linkage judgment index system refers to a multi-dimensional evaluation framework that integrates the target orientation deviation threshold, vibration signal change threshold, and trajectory execution progress threshold. It is used to comprehensively determine whether the current tracking trajectory is suitable for the target movement state and terrain features. It is constructed by associating the target movement orientation, terrain features, and tracking trajectory execution degree. The target orientation deviation threshold is used to define the critical value at which the target has exceeded the current tracking trajectory's field of view and the trajectory needs to be forcibly updated; The vibration signal change threshold is used to reflect the degree of mismatch between the current tracking trajectory and the actual terrain features; The trajectory execution progress threshold is used to determine the lifecycle stage of the currently tracked trajectory; Collect target location and vibration data, and record the distance ratio between the already tracked trajectory and the currently planned tracking trajectory. Through a linkage judgment indicator system, determine the trigger for the trajectory update mechanism.

10. An all-terrain maneuvering target tracking system, used to implement the all-terrain maneuvering target tracking method according to any one of claims 1-9, characterized in that, include: The module includes: orientation update module, range delineation module, terrain construction module, region division module, trajectory optimization module, balance adjustment module, and update judgment module. The orientation update module is used to track the target's orientation and adjust the direction of travel in real time; the range delineation module is used to dynamically delineate the acquisition range; the terrain construction module is used to control the camera to acquire and process images to construct a topographic map; the region division module is used to preprocess the topographic map and extract landform features, construct a comprehensive feature matrix, and divide the passable area; the trajectory optimization module is used to generate and optimize the tracking trajectory in the passable area. The balance adjustment module is used to adaptively adjust the balance component parameters based on the terrain features of the tracking trajectory area; the update judgment module is used to determine whether to trigger a trajectory update by comprehensively considering the vibration signal, trajectory execution progress, and target orientation change.

Images