A landing safety zone autonomous rapid search method, system, device and medium based on multiple terrain features and quadtree

Abstract

The application discloses a kind of based on the autonomous fast search method, system, equipment and medium of landing safety zone of multiple terrain features and quadtree, comprising: obtaining landing terrain point cloud data, constructs terrain grid map and calculates slope, roughness, plane fitting elevation difference, curvature and other basic terrain features, constructs integrated risk index after normalization and generates multi-band feature map by fusion;Multi-scale segmentation algorithm is used to carry out regional segmentation to multi-band feature map, and binary classification is generated by combining classification model to generate obstacle binary graph;Based on the coarse screening locking candidate partition of obstacle binary graph using quadtree structure, the center of the largest continuous safety area is calculated as the center of the largest inscribed circle using edge submersion algorithm in candidate partition, complete the autonomous fast search of landing safety zone based on multiple terrain features and quadtree.The application significantly shortens the search time under the premise of ensuring the accuracy of safety zone selection, and balances the reliability and efficiency of landing area selection.

Description

Technical Field

[0001] This invention relates to the field of precise and safe landing technology for planetary exploration, and in particular to a method, system, device and medium for autonomous and rapid search of landing safety zones based on multiple terrain features and quadtrees. Background Technology

[0002] The complex and dangerous surfaces of extraterrestrial bodies, riddled with obstacles such as craters and rocky slopes, present a significant challenge to the autonomous and safe landing of probes due to the lack of detailed topographic information and the inability to anticipate small-scale obstacles that threaten landing safety. For example, Intuitive Machines' Odysseus and Athena probes, iSpace's HAKUTO-R probe, and the Japan Aerospace Exploration Agency's (JAXA) Smart Lander for Investigating Moon (SLIM) probe all failed to land smoothly on the lunar surface.

[0003] my country's Chang'e 3, 4, 5, and 6 lunar probes and the Tianwen-1 Mars lander used laser 3D imaging sensors to acquire 3D point cloud data of the terrain below the lander at an altitude of approximately 100 meters above the lunar and Martian surfaces. This data was used to accurately detect obstacles such as craters, rocks, and slopes that could threaten landing safety, and to select a safe landing area, enabling the lander to avoid obstacles and land safely. Due to fuel limitations, the need to quickly select a safe landing site after acquiring the 3D point cloud data of the landing area presented a significant challenge for the onboard computer with its limited computing power. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an autonomous and rapid search method, system, device, and medium for landing safety zones based on multiple terrain features and quadtrees to solve the problem that existing extraterrestrial landing missions have difficulty in quickly selecting safe landing sites after acquiring point clouds.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides an autonomous and rapid method for searching landing safety zones based on multiple terrain features and quadtrees, including: Acquire landing terrain point cloud data, construct a terrain raster map, calculate multiple basic terrain features corresponding to each raster, normalize the basic terrain features, construct a comprehensive hazard index based on the processed features, and fuse the basic terrain features with the comprehensive hazard index to generate a multi-band feature map. The multi-scale segmentation algorithm is used to segment the multi-band feature map into regions, and the segmented regions are classified into two categories by a classification model to generate a binary obstacle map. The obstacle binary map is coarsely screened based on a spatial recursive partitioning structure to lock candidate partitions. The edge flooding algorithm is then used to calculate the center of the largest inscribed circle within the candidate partitions. The center of the largest inscribed circle is then used as the center of the largest continuous safe area to complete the autonomous and rapid search of the landing safe area.

[0007] As a preferred embodiment of the autonomous and rapid landing safety zone search method based on multi-terrain features and quadtrees described in this invention, the step of calculating the center of the largest inscribed circle within the candidate partition using the edge submergence algorithm includes: Within the candidate partition, the region consisting of all safe pixels is taken as the initial safe region; Using the obstacle pixel as a seed, iterative erosion is performed on the initial safe region in the eight-neighbor direction (i.e., the obstacle pixel is gradually submerged and expanded to the adjacent safe pixel). Record the area change data of the remaining safe pixel set after each iteration; The geometric center of the remaining safe pixel region in the last iteration with an area greater than zero is determined as the center of the maximum inscribed circle.

[0008] As a preferred embodiment of the autonomous and rapid landing safety zone search method based on multi-terrain features and quadtrees described in this invention, the step of coarsely screening the obstacle binary map based on a spatial recursive partitioning structure to lock candidate partitions includes: The obstacle binary graph is recursively partitioned using a quadtree structure to generate a tree-like data structure containing multiple levels of nodes, wherein the end of the tree-like data structure is a leaf node. Traverse the leaf nodes, calculate the number of obstacle pixels inside each leaf node, and filter out target leaf nodes whose number of obstacle pixels is lower than a preset tolerance; The spatial region covered by the target leaf node is extracted and marked as the candidate partition.

[0009] The beneficial effects of this preferred technical solution are that by using a quadtree structure for recursive partitioning and filtering, multi-resolution hierarchical management of the obstacle binary map is achieved, which can quickly eliminate large-area dangerous areas and lock potential safe areas, thereby significantly reducing the computational range of subsequent precise searches, effectively reducing computational complexity and improving the real-time performance of landing area selection.

[0010] As a preferred embodiment of the autonomous and rapid landing safety zone search method based on multi-terrain features and quadtrees described in this invention, the step of segmenting the multi-band feature map using a multi-scale segmentation algorithm includes: Extract the feature vector of each grid in the multi-band feature map to construct the feature space; In the feature space, the feature heterogeneity between adjacent grid cells is calculated using an iterative merging algorithm, and adjacent grid cells with feature heterogeneity below a preset threshold are aggregated. As the number of iterations increases, the aggregated raster set is defined as an independent segmentation object, generating multiple homogeneous regions with consistent internal features.

[0011] As a preferred embodiment of the autonomous and rapid landing safety zone search method based on multi-terrain features and quadtrees described in this invention, wherein: the construction of a comprehensive hazard index based on the processed features includes: Assign corresponding weight coefficients to each normalized basic terrain feature to obtain a weighted feature set; A linear superposition operation is performed on each feature value in the weighted feature set to obtain a fused feature value; The fused feature value is used as a quantitative representation of the comprehensive hazard index to generate the comprehensive hazard index.

[0012] The beneficial effect of this preferred technical solution is that by assigning weights to different features and performing linear fusion, multi-dimensional terrain features are transformed into a single quantitative comprehensive hazard index. While retaining the key information of each feature, it achieves a unified representation of the terrain hazard level, providing an efficient and intuitive data foundation for the subsequent rapid identification of safe landing areas.

[0013] As a preferred embodiment of the autonomous and rapid landing safety zone search method based on multi-terrain features and quadtrees described in this invention, the normalization processing of the basic terrain features includes: For each of the basic terrain features, the maximum and minimum values ​​of the basic terrain features are extracted, and a linear transformation function is constructed. The original feature values ​​of the basic terrain features are scaled using the linear transformation function, and the feature values ​​are mapped to a preset uniform numerical range to obtain normalized feature data.

[0014] As a preferred embodiment of the autonomous and rapid search method for landing safety zones based on multiple terrain features and quadtrees described in this invention, the multiple basic terrain features include: slope, roughness, plane fitting elevation difference, and curvature.

[0015] Secondly, the present invention provides an autonomous and rapid landing safety zone search system based on multiple terrain features and quadtrees, comprising: The multi-band feature construction module is used to acquire landing terrain point cloud data, construct a terrain raster map, calculate multiple basic terrain features corresponding to each raster, normalize the basic terrain features, construct a comprehensive hazard index based on the processed features, and fuse the basic terrain features with the comprehensive hazard index to generate a multi-band feature map. The obstacle region segmentation and recognition module is used to segment the multi-band feature map using a multi-scale segmentation algorithm, and to perform binary classification on the segmented regions using a classification model to generate an obstacle binary map. The safe landing zone fast search module is used to coarsely screen the obstacle binary map based on the spatial recursive partitioning structure to lock candidate partitions, and use the edge flooding algorithm to calculate the center of the largest inscribed circle in the candidate partitions. The center of the largest inscribed circle is used as the center of the largest continuous safe area to complete the autonomous fast search of the landing safe area.

[0016] Thirdly, the present invention provides an electronic device, comprising: Memory, used to store programs; A processor is configured to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the autonomous and rapid landing safety zone search method based on multi-terrain features and quadtrees.

[0017] Fourthly, the present invention provides a computer-readable storage medium, comprising: when the program is executed by a processor, the steps of implementing the autonomous and rapid landing safety zone search method based on multiple terrain features and quadtree.

[0018] The beneficial effects of this invention are as follows: By recursively partitioning the binary obstacle map using a quadtree structure and generating a tree-like data structure, this invention achieves multi-resolution hierarchical management of hazardous terrain areas, constructing a coarse-to-fine search foundation; by traversing leaf nodes and calculating the number of hazardous pixels within them to filter target leaf nodes below a preset tolerance, it achieves rapid identification of potential safe areas and effectively eliminates interference from large areas of high-risk regions; by extracting and marking the spatial area covered by the target leaf node as a candidate partition, it achieves precise locking of the search range, significantly reducing the computational domain for subsequent precise calculations, thereby greatly reducing algorithm complexity and improving the real-time performance of landing decisions. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 A schematic diagram of the basic process of an autonomous and rapid landing safety zone search method based on multiple terrain features and quadtrees, provided as an embodiment of the present invention; Figure 2 This is an example diagram illustrating an autonomous and rapid landing safety zone search method based on multiple terrain features and quadtrees, provided as an embodiment of the present invention. Detailed Implementation

[0020] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0021] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for autonomous and rapid search of landing safety zones based on multiple terrain features and quadtrees is provided, including: S100: Acquire landing terrain point cloud data, construct a terrain raster map, calculate multiple basic terrain features corresponding to each raster, normalize the basic terrain features, construct a comprehensive hazard index based on the processed features, and fuse the basic terrain features with the comprehensive hazard index to generate a multi-band feature map. S200: Use a multi-scale segmentation algorithm to segment the multi-band feature map into regions, and combine it with a classification model to perform binary classification on the segmented regions to generate a binary obstacle map; S300: Based on the spatial recursive partitioning structure, the obstacle binary map is coarsely screened to lock candidate partitions, and the edge flooding algorithm is used to calculate the center of the largest inscribed circle in the candidate partition. The center of the largest inscribed circle is used as the center of the largest continuous safe area to complete the autonomous and fast search of the landing safe area.

[0022] It should be noted that extraterrestrial landing missions involve a complex process of high-speed entry, descent, and landing. The varying and often unknown parameters of extraterrestrial atmospheres, gravity environments, and terrain present significant challenges to the lander's deceleration strategies, thermal protection design, and autonomous control. Furthermore, the vast distances to extraterrestrial bodies result in significant communication delays, preventing real-time intervention by ground teams. The lander must possess highly autonomous "perception-decision-control" capabilities to handle unforeseen circumstances. Simultaneously, the complex obstacles on extraterrestrial surfaces, such as craters and rocks, place extremely high demands on landing accuracy and obstacle avoidance capabilities. Therefore, autonomous and safe landing and precise obstacle avoidance technologies for extraterrestrial bodies are crucial.

[0023] Therefore, addressing the aforementioned problem of existing extraterrestrial landing missions struggling to quickly select safe landing sites after acquiring point clouds, this method constructs a comprehensive hazard index based on multiple terrain features through steps S100-S300. It then utilizes multi-scale segmentation and random forest classification techniques, combined with quadtree and edge flooding methods, to accurately and rapidly select safe landing sites. Compared to traditional methods, this approach offers higher efficiency and safety in safe zone selection. This method can be used for deep space probes to rapidly select safe landing sites on the surfaces of extraterrestrial bodies with complex and unknown topography and diverse environments.

[0024] Example 2, this is an embodiment of the present invention, which provides an autonomous and rapid landing safety zone search method based on multiple terrain features and quadtrees, based on the previous embodiment, including: In this embodiment of the application, the acquisition of landing terrain point cloud data in step S100 specifically involves: processing the laser three-dimensional point cloud acquired by the landing laser three-dimensional sensor; multiple basic terrain features, including: slope S, roughness R, plane fitting elevation difference E, and curvature D.

[0025] In this embodiment of the application, step S100, which normalizes the basic terrain features, includes: normalizing the values ​​of each basic terrain feature using a min-max normalization method to obtain normalized basic feature values. For example, for slope S, roughness R, plane fitting elevation difference, and curvature D, the normalized values ​​are obtained. , , , .

[0026] In an optional implementation, the multiple basic terrain features mentioned in step S100 may further include complexity features based on normal vector statistics: for the laser three-dimensional point cloud acquired by the landing laser three-dimensional sensor, the normal vector is calculated point by point, and based on the distribution variance of the normal vector in the local neighborhood or the rate of change of the angle between the normal vectors, a basic feature characterizing the terrain complexity is constructed; this feature, as one of the basic terrain features, participates in the subsequent normalization processing and the weighted fusion of comprehensive hazard indicators.

[0027] In an optional implementation, the multiple basic terrain features mentioned in step S100 may further include fractal dimension features: for the laser three-dimensional point cloud, the fractal dimension in the local area is calculated using box counting or variational method to characterize the irregularity and complexity of the terrain; this fractal dimension, as one of the basic terrain features, participates in normalization processing, construction of comprehensive hazard indicators, and generation of multi-band feature maps together with features such as slope, roughness, plane fitting elevation difference, and curvature.

[0028] In this embodiment of the application, step S100, which constructs a comprehensive hazard index based on the processed features, includes: Assign corresponding weight coefficients to each normalized basic terrain feature to obtain a weighted feature set; A linear superposition operation is performed on the feature values ​​in the weighted feature set to obtain the fused feature value; The fusion feature value is used as a quantitative representation of the comprehensive hazard index to generate the comprehensive hazard index.

[0029] Specifically, the comprehensive risk index I is calculated using formula (1): (1) in, , , , These are the weight coefficients corresponding to each normalized basic feature. These weights reflect the relative importance of each basic indicator in terrain landing hazard assessment. The weights are usually determined based on the actual engineering capability constraints of the target probe. Weights can be set using historical lunar landing data or ground simulation experiments to improve the accuracy of hazard indicators in assessing landing risk. The comprehensive hazard indicator I is then normalized using the min-max normalization method. .

[0030] In this embodiment of the application, in step S100, the normalized basic terrain feature maps calculated in step S100 are fused with the normalized comprehensive hazard index map into a multi-band feature map, the number of which is the sum of the number of basic terrain features used and 1; for example, when only four basic features are used—slope, roughness, plane fitting elevation difference, and curvature—a 5-band feature map is generated. ; In this embodiment of the application, step S200 uses a multi-scale segmentation algorithm to perform region segmentation on the multi-band feature map, including: Extract the feature vector of each grid in the multi-band feature map to construct the feature space; In the feature space, the feature heterogeneity between adjacent grid cells is calculated using an iterative merging algorithm, and adjacent grid cells with feature heterogeneity below a preset threshold are aggregated. As the number of iterations increases, the aggregated raster set is defined as an independent segmentation object, generating multiple homogeneous regions with consistent internal features.

[0031] In this embodiment of the application, basic terrain features are fused with comprehensive hazard indicators to generate a multi-band feature map, including: Multiple basic terrain features after normalization are treated as independent channels, and the comprehensive hazard index is treated as an additional channel. These are stacked at the pixel level to construct a multi-band feature map with multiple channel dimensions.

[0032] In this embodiment of the application, step S200 combines a classification model to perform binary classification on the segmented regions, including: for the K segmented regions, a random forest classification model is used to perform binary classification on each region. If the region belongs to a dangerous region, it is assigned a value of 1, and if it belongs to a safe region, it is assigned a value of 0, thereby obtaining a binary obstacle map.

[0033] In an optional implementation, the method for generating the obstacle binary map in step S200 can also directly input the multi-band feature map generated in step S100 into a pre-trained semantic segmentation network, and directly output a pixel-level prediction map containing the probability that each grid belongs to an obstacle or a safe area through the forward inference of the network, and use the prediction map as the obstacle binary map after thresholding.

[0034] In an optional implementation, the method for generating the obstacle binary map in step S200 can also automatically calculate the optimal segmentation threshold using an adaptive threshold algorithm on the multi-band feature map or comprehensive hazard index map generated in step S100, compare the hazard index value of each grid in the map with the threshold, mark areas above the threshold as dangerous, and mark areas below the threshold as safe, thereby directly generating the obstacle binary map.

[0035] In this embodiment of the application, the segmented regions are classified in binary order using a classification model to generate a binary obstacle map, including: The classification model is used to output a category label for each region, which includes safety and danger categories. Regions marked as safe are assigned the first pixel value in the obstacle binary map, and regions marked as hazardous are assigned the second pixel value.

[0036] In this embodiment of the application, step S300, which involves coarsely screening the obstacle binary map based on a spatial recursive partitioning structure to lock down candidate partitions, includes: A quadtree structure is used to recursively partition the obstacle binary map. The entire obstacle binary map is taken as the root node, and the current node is divided into four child nodes each time. The process is repeated layer by layer to generate a tree data structure containing multiple levels of nodes. When a node meets a preset termination condition, the further division of that node is stopped. The preset termination condition includes any one or a combination of the following: the recursion depth of the node reaches a preset maximum iteration depth, the node size reaches a preset minimum allowable size, or all pixels in the node are safe pixels or all pixels are obstacle pixels. After the recursion terminates, the node at the end of the tree data structure is taken as the leaf node. All leaf nodes are traversed and the number of obstacle pixels inside each leaf node is calculated. Leaf nodes with the number of obstacle pixels lower than the preset tolerance are selected as target leaf nodes. The spatial region covered by the target leaf node is extracted and marked as a candidate partition.

[0037] In an optional implementation, the coarse screening method for the safe area in step S300 can also set up a variety of sliding windows of different sizes, perform traversal scanning on the obstacle binary map with a fixed step size, calculate the obstacle pixel density within the coverage area of ​​each window, and mark and extract the areas corresponding to windows with obstacle pixel density lower than a preset tolerance as candidate partitions.

[0038] In an optional implementation, the coarse screening method for safe regions in step S300 can also perform connected component labeling on the safe pixels in the obstacle binary map to identify all connected safe region patches, calculate the area of ​​each connected region and remove patches with an area smaller than a preset threshold, and use the remaining large-area connected regions as candidate partitions.

[0039] In this embodiment of the application, step S300, which uses the edge flooding algorithm to calculate the center of the largest inscribed circle within the candidate partition, includes: For the candidate partition output in step S300, the region formed by all safe pixels inside it is taken as the initial safe region. Using the obstacle pixel as the seed, iterative erosion is performed on the initial safe region in the eight-neighbor direction (i.e., the obstacle pixel gradually floods and expands to the adjacent safe pixel). For each extended barrier pixel, continue to extend the erosion of the safe pixel to its eight neighbors, and repeat the above iterative erosion process; Record the area change data of the remaining safe pixel set after each iteration; The geometric center of the remaining safe pixel region in the last iteration with an area greater than zero is determined as the center of the largest inscribed circle.

[0040] In an optional implementation, the calculation of the center of the largest inscribed circle within the candidate partition in step S300 can also be performed as follows: perform Euclidean distance transformation on the candidate partition, calculate the straight-line distance from each safe pixel to the nearest obstacle pixel within the candidate partition, and generate a distance map; determine the pixel with the largest value in the distance map as the center of the largest inscribed circle.

[0041] In an optional implementation, the calculation of the center of the maximum inscribed circle within the candidate partition in step S300 can also be performed as follows: extract the set of intersection points between obstacle pixels and safe pixels within the candidate partition, construct a Voronoi diagram based on the set of points, filter out the Voronoi vertices located inside the safe pixel region, calculate the distance from each vertex to the nearest obstacle pixel edge point, and determine the vertex with the largest distance value as the center of the maximum inscribed circle.

[0042] In the embodiments of this application, by adopting the above steps, it is possible to accurately and quickly select the landing safety zone by using multi-scale segmentation and random forest classification technology, combined with quadtree and edge flooding methods, while taking into account various terrain risk characteristics, thus providing support for the accurate and safe landing of extraterrestrial bodies.

[0043] Example 3, referring to Figure 2 Table 1 illustrates one embodiment of the present invention, which provides an autonomous and rapid landing safety zone search method based on multiple terrain features and quadtrees. To verify the beneficial effects of the present invention, specific implementation methods and their effects are scientifically demonstrated.

[0044] The specific details of this embodiment are as follows: I. Experimental Environment Hardware configuration: A computing platform equipped with an AMD Ryzen 9 5900HS 3.30GHz CPU and 32.0GB of memory.

[0045] Data source: Laser 3D point cloud data with an area of ​​approximately 50 meters × 50 meters, acquired at a height of approximately 100 meters at the Tongji University Lunar and Deep Space Exploration Precision Mapping Comprehensive Experimental Field (e.g., Figure 2 (as shown in (a)).

[0046] II. Experimental Procedure 1. Laser 3D point cloud preprocessing and feature extraction The acquired laser 3D point cloud data is preprocessed by denoising, filtering, and other operations.

[0047] Single terrain features such as slope (S), roughness (R), elevation difference (E), and curvature (D) are extracted from the preprocessed point cloud data.

[0048] Construct a comprehensive hazard index (I) that integrates multiple terrain features; the formula is as follows: (in , , , The weighting coefficients for each feature are determined based on the actual terrain hazard assessment requirements; and the comprehensive hazard index I is normalized using the min-max normalization method to obtain the normalized comprehensive hazard index.

[0049] 2. Obstacle Binary Map Generation The multi-band feature map generated in step S100 is segmented using a multi-scale segmentation method to obtain a segmentation result image (e.g., Figure 2 (b) shown).

[0050] A random forest classification model is used to classify each segmented region into two categories (obstacles / non-obstacles), generating a binary obstacle map, where obstacle regions are marked as 1 and non-obstacle (safe) regions are marked as 0.

[0051] 3. Coarse screening of safe regions (quadtree search) A quadtree structure is constructed using a binary obstacle map as input.

[0052] The quadtree nodes are recursively divided, and obstacle pixels are detected in the area covered by each node. Nodes with fewer than the preset tolerance of obstacle pixels in their leaf nodes are retained as candidate partitions.

[0053] 4. Maximum safety zone location (edge ​​flooding method) For each candidate partition, the region consisting of all safe pixels within it is taken as the initial safe region.

[0054] An edge flooding algorithm is used, with obstacle pixels in the candidate partition as seeds, and iterative erosion is performed on the initial safe region in the direction of eight neighborhoods (i.e. obstacle pixels are gradually flooded and expanded to adjacent safe pixels), and the area of ​​the remaining safe pixel set after each iteration is recorded.

[0055] The geometric center of the remaining safe pixel region in the last iteration with an area greater than zero is determined as the center of the maximum inscribed circle; the product of the number of this last iteration and the pixel size is the radius of the maximum safe landing area (this radius characterizes the scale of the maximum continuous safe area and is used as a reference for subsequent landing decisions), thus obtaining the maximum safe landing area (e.g., ...). Figure 2 (c) shows the red circular area.

[0056] III. Experimental Results 1. Obstacle detection performance The obstacle detection rates of single features (slope S, roughness R, elevation difference E, curvature D) and comprehensive hazard index (I) were compared, and the results are shown in Table 1. The comprehensive hazard index had a false negative rate of 1 and a false negative rate of 0, with an accuracy of 98%, a recall of 98%, and a precision of 100%, which is significantly better than the single feature method, proving that the comprehensive hazard index can more accurately identify terrain obstacles.

[0057] Table 1. Comparison of obstacle detection rates using a single feature and the comprehensive hazard index constructed using this invention.

[0058] from Figure 2 (c) It can be seen that the final safe landing area was located in a flat area, successfully avoiding obstacles such as pits, rocks and steep slopes in the site, verifying the effectiveness of the method in obstacle avoidance.

[0059] 2. Efficiency of selecting safe zones Using only the quadtree search method takes 0.21 seconds, but the selection of the safe zone is poor (the maximum safe zone is not accurately located); using the edge flooding method directly takes 0.63 seconds and the effect is better; using the quadtree-edge flooding method of the present invention takes a total of 0.33 seconds, which ensures both selection accuracy and high efficiency.

[0060] Reasons for efficiency improvement: Quadtree search quickly narrows down the search range of the safe zone, reducing the computational cost of the edge flooding method; the edge flooding method accurately locates the maximum safe zone within the candidate partition, ensuring the accuracy of the results.

[0061] Example 4 is an embodiment of the present invention. This embodiment differs from the first embodiment in that it provides an autonomous and rapid landing safety zone search system based on multiple terrain features and quadtrees.

[0062] It should be noted that the technical solution of the autonomous rapid search system for landing safety zones based on multi-terrain features and quadtrees is based on the same concept as the above-mentioned autonomous rapid search method for landing safety zones based on multi-terrain features and quadtrees. For details not described in detail in the technical solution of the autonomous rapid search system for landing safety zones based on multi-terrain features and quadtrees in this embodiment, please refer to the description of the above-mentioned autonomous rapid search method for landing safety zones based on multi-terrain features and quadtrees.

[0063] This embodiment presents an autonomous and rapid landing safety zone search system based on multiple terrain features and quadtrees, comprising: The multi-band feature construction module is used to acquire landing terrain point cloud data, construct terrain raster maps, calculate multiple basic terrain features corresponding to each raster, normalize the basic terrain features, construct a comprehensive hazard index based on the processed features, and fuse the basic terrain features with the comprehensive hazard index to generate a multi-band feature map. The obstacle region segmentation and recognition module is used to segment multi-band feature maps using a multi-scale segmentation algorithm, and then perform binary classification on the segmented regions using a classification model to generate a binary obstacle map. The safe landing zone fast search module is used to coarsely screen the obstacle binary map based on the spatial recursive partitioning structure to lock candidate partitions, and use the edge flooding algorithm to calculate the center of the largest inscribed circle in the candidate partitions. The center of the largest inscribed circle is used as the center of the largest continuous safe area to complete the autonomous and fast search of the landing safe area.

[0064] This embodiment also provides an electronic device applicable to a method for autonomous and rapid search of landing safety zones based on multiple terrain features and quadtrees, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement an autonomous and rapid landing safety zone search method based on multiple terrain features and a quadtree, as proposed in the above embodiments.

[0065] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements an autonomous and rapid landing safety zone search method based on multiple terrain features and quadtrees, as proposed in the above embodiments.

[0066] The storage medium proposed in this embodiment and the method for autonomous and rapid search of landing safety zones based on multiple terrain features and quadtree proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0067] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0068] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for autonomous and rapid search of landing safety zones based on multiple terrain features and quadtrees, characterized in that, include: Acquire landing terrain point cloud data, construct a terrain raster map, calculate multiple basic terrain features corresponding to each raster, normalize the basic terrain features, construct a comprehensive hazard index based on the processed features, and fuse the basic terrain features with the comprehensive hazard index to generate a multi-band feature map. The multi-scale segmentation algorithm is used to segment the multi-band feature map into regions, and the segmented regions are classified into two categories by a classification model to generate a binary obstacle map. The obstacle binary map is coarsely screened based on a spatial recursive partitioning structure to lock candidate partitions. The edge flooding algorithm is then used to calculate the center of the largest inscribed circle within the candidate partitions. The center of the largest inscribed circle is then used as the center of the largest continuous safe area to complete the autonomous and rapid search of the landing safe area.

2. The autonomous and rapid landing safety zone search method based on multiple terrain features and quadtrees as described in claim 1, characterized in that: The step of using the edge-submerging algorithm to calculate the center of the largest inscribed circle within the candidate partition includes: Within the candidate partition, the region consisting of all safe pixels is taken as the initial safe region; Using obstacle pixels as seeds, iterative erosion is performed on the initial safe region in the eight-neighbor direction, that is, the obstacle pixels are gradually submerged and expanded to adjacent safe pixels. For each extended barrier pixel, continue to extend the erosion of the safe pixel to its eight neighbors, and repeat the above iterative erosion process; Record the area change data of the remaining safe pixel set after each iteration; The geometric center of the remaining safe pixel region in the last iteration with an area greater than zero is determined as the center of the maximum inscribed circle.

3. The autonomous and rapid landing safety zone search method based on multiple terrain features and quadtrees as described in claim 1 or 2, characterized in that: The method of coarsely screening the obstacle binary map based on a spatial recursive partitioning structure to narrow down candidate partitions includes: The obstacle binary map is recursively partitioned using a quadtree structure. The entire obstacle binary map is taken as the root node, and each time the current node is divided into four child nodes. This process is repeated layer by layer downwards to generate a tree-like data structure containing multiple levels of nodes. When a node meets a preset termination condition, the partitioning of that node stops. The preset termination condition includes any one or a combination of the following: the recursion depth of the node reaches a preset maximum iteration depth, the node size reaches a preset minimum allowable size, or all pixels in the node are safe pixels or all pixels are obstacle pixels. After the recursion terminates, the node at the end of the tree-like data structure is taken as a leaf node. Traverse the leaf nodes, calculate the number of obstacle pixels inside each leaf node, and filter out target leaf nodes whose number of obstacle pixels is lower than a preset tolerance; The spatial region covered by the target leaf node is extracted and marked as the candidate partition.

4. The autonomous and rapid landing safety zone search method based on multiple terrain features and quadtrees as described in claim 3, characterized in that: The process of segmenting the multi-band feature map using a multi-scale segmentation algorithm includes: Extract the feature vector of each grid in the multi-band feature map to construct the feature space; In the feature space, the feature heterogeneity between adjacent grid cells is calculated using an iterative merging algorithm, and adjacent grid cells with feature heterogeneity below a preset threshold are aggregated. As the number of iterations increases, the aggregated raster set is defined as an independent segmentation object, generating multiple homogeneous regions with consistent internal features.

5. The autonomous and rapid landing safety zone search method based on multiple terrain features and quadtrees as described in claim 4, characterized in that: The construction of a comprehensive hazard index based on the processed features includes: Assign corresponding weight coefficients to each normalized basic terrain feature to obtain a weighted feature set; A linear superposition operation is performed on each feature value in the weighted feature set to obtain a fused feature value; The fused feature value is used as a quantitative representation of the comprehensive hazard index to generate the comprehensive hazard index.

6. The autonomous and rapid landing safety zone search method based on multiple terrain features and quadtrees as described in claim 5, characterized in that: The normalization process for the basic terrain features includes: For each of the basic terrain features, the maximum and minimum values ​​of the basic terrain features are extracted, and a linear transformation function is constructed. The original feature values ​​of the basic terrain features are scaled using the linear transformation function, and the feature values ​​are mapped to a preset uniform numerical range to obtain normalized feature data.

7. The autonomous and rapid landing safety zone search method based on multiple terrain features and quadtrees as described in claim 6, characterized in that: The various basic terrain features include: slope, roughness, plane fitting elevation difference, and curvature.

8. An autonomous and rapid landing safety zone search system based on multiple terrain features and quadtrees, employing the method described in any one of claims 1-7, characterized in that, include: The multi-band feature construction module is used to acquire landing terrain point cloud data, construct a terrain raster map, calculate multiple basic terrain features corresponding to each raster, normalize the basic terrain features, construct a comprehensive hazard index based on the processed features, and fuse the basic terrain features with the comprehensive hazard index to generate a multi-band feature map. The obstacle region segmentation and recognition module is used to segment the multi-band feature map using a multi-scale segmentation algorithm, and to perform binary classification on the segmented regions using a classification model to generate an obstacle binary map. The safe landing zone fast search module is used to coarsely screen the obstacle binary map based on the spatial recursive partitioning structure to lock candidate partitions, and use the edge flooding algorithm to calculate the center of the largest inscribed circle in the candidate partitions. The center of the largest inscribed circle is used as the center of the largest continuous safe area to complete the autonomous fast search of the landing safe area.

9. An electronic device, characterized in that, include: Memory, used to store programs; A processor for executing the program, wherein the program, when executed by the processor, implements the steps of the method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a program, characterized in that, When the program is executed by a processor, it implements the steps of the method as described in any one of claims 1-7.

Images