Region exploration methods, systems, devices, and storage media based on point cloud clustering

Abstract

The present invention proposes a region exploration method, system, device and storage medium based on point cloud clustering, wherein the method comprises the following steps: generating a viewpoint queue by a viewpoint generation algorithm according to input point cloud data; dividing the viewpoint queue into different clusters by viewpoint clustering to obtain a cluster queue; calculating the current position cluster c in real time j1 The cost function between the current clusters includes a boundary heuristic function, a dynamic loss cost function and a distance cost function; according to the current exploration device information, the cluster point with the smallest cost value is iteratively selected as the next expansion area; the process is repeated to guide the exploration device to explore the unknown area in real time. The present invention can solve the problems of high computational complexity, unreasonable exploration path, repeated access and low efficiency in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of robot environment exploration, and in particular to a region exploration method, system, device and storage medium based on point cloud clustering. Background Art

[0002] In recent years, with the rapid development of robots, autonomous vehicles and drones, 3D space reconstruction technology has been increasingly used in various scenarios. In order to accurately model unknown environments, point cloud data has become the main data form for 3D reconstruction. Point cloud data is acquired through sensors such as LiDAR, structured light, and stereo vision, and can provide the precise position of the surface of an object in 3D space. How to efficiently and accurately process these point cloud data to achieve complete exploration of the target area has become one of the current research hotspots.

[0003] In the prior art, regional exploration methods are usually based on simple path planning algorithms, such as A*, Dijkstra algorithm, etc., or methods based on gradient descent. These methods mainly rely on traditional space segmentation strategies or viewpoint-based path planning to achieve environmental exploration. However, when dealing with complex environments, especially in high-dimensional spaces, these methods often exhibit problems such as high computational complexity and inability to guarantee a global optimal solution. Therefore, the present invention proposes a new regional exploration method, which can optimize the exploration order based on effective clustering of point cloud data, thereby achieving efficient and comprehensive exploration of unknown areas. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention proposes a regional exploration method, system, device and storage medium based on point cloud clustering, which can solve the problems of high computational complexity, unreasonable exploration path, repeated access and low efficiency in the prior art.

[0005] The technical solution of the present invention is achieved in this way:

[0006] In one aspect, the present invention provides a region exploration method based on point cloud clustering, comprising the following steps:

[0007] According to the input point cloud data, the viewpoint queue L is generated through the viewpoint generation algorithm view ={p1,p2,...,p i ,...,p n}, where viewpoint p i Belongs to the three-dimensional viewpoint;

[0008] The above viewpoint queues are divided into different clusters c by viewpoint clustering j , to obtain the clustering queue L clusters ={c1,c2,...,c j,...,c m};

[0009] Real-time computing explores the current location cluster of the device j1 A cost function between the clusters and other current clusters, wherein the cost function includes a boundary heuristic function, a dynamic loss cost function and a distance cost function;

[0010] According to the current exploration device information, the cluster point with the smallest cost is selected as the next expansion area; this process is repeated to guide the exploration device to explore the unknown area in real time.

[0011] Preferably, the step of generating a viewpoint queue by a viewpoint generation algorithm according to the input point cloud data further includes the following steps:

[0012] The validity of the point cloud is detected by incrementally constructing a point cloud map through lidar, and the valid point cloud data is input into the viewpoint generation algorithm to generate a viewpoint queue.

[0013] Preferably, the step of dividing the above viewpoint queues into different clusters by viewpoint clustering to obtain cluster queues includes:

[0014] The current exploration device location is the cluster center cluster j , determine all unclustered viewpoints p i and cluster center cluster j Is the maximum cluster distance between them less than the preset cluster distance d? cluster ;

[0015] Through the unclustered viewpoint p i and cluster center cluster j The connection detection function between them determines the unclustered viewpoints p i and cluster center cluster j Whether there is a collision between them;

[0016] When the unclustered viewpoint p i and cluster center cluster j The maximum cluster distance between them is less than the preset cluster distance d cluster And the unclustered viewpoint p i and cluster center cluster j When there is no collision between the cluster viewpoints p i Join the cluster c j Otherwise, p i Create a new cluster for the center;

[0017] Repeat the above process to obtain the cluster queue L clusters ={c1,c2,...,c j ,...,cm}, where each cluster c j Contains all viewpoints p that meet the clustering condition i .

[0018] Preferably, the cost function is f total (c j1 ,c j2 ) is:

[0019] f total (c j1 ,c j2 )=αf b (c j1 ,c j2 )+βf dy (c j1 ,c j2 )+γf d (c j1 ,c j2 ),

[0020] Among them, α, β, γ represent the weight ratio, f b (c j1 ,c j2 ) is the boundary heuristic function, f dy (c j1 ,c j2 ) is the dynamic loss cost function, f d (c j1 ,c j2 ) is the distance cost function.

[0021] Preferably, the boundary heuristic function f b (c j1 ,c j2 ) value is determined by the following formula:

[0022] f b (c j1 ,c j2 )=GetBarrierDis(c j2 )-GetBarrierDis(c j1 )

[0023] GetBarrierDis(c j1 )=min(|c j1,x_center -x barrier |,|c j1,y_center -y barrier |)

[0024] where c j1,x_center 、c j1,y_center Represents cluster c j1The average coordinate value of the viewpoint in two dimensions, x barrier ,y barrier Indicates the boundary value of the search area, GetBarrierDis(c j1 ) Calculate the distance plane c j1 Distance from the border.

[0025] Preferably, the dynamic loss cost function f dy (c j1 ,c j2 ) value is determined by the following formula:

[0026]

[0027] where c j1,center For cluster c j1 The average position of all viewpoints in j2,center For cluster c j2 The average position of all viewpoints in j1,θ For cluster c j1 The attitude angles of all viewpoints in j2,θ For cluster c j2 The pose angles of all viewpoints in .

[0028] Preferably, the distance cost function f d (c j1 ,c j2 ) value is determined by the following formula:

[0029] f d (c j1 ,c j2 )=||c j1,center -c j2,center ||

[0030] where c j1,center For cluster c j1 The average position of all viewpoints in j2,center For cluster c j2 The average position of all viewpoints in .

[0031] On the other hand, the present invention also provides a region exploration system based on point cloud clustering, characterized in that it includes:

[0032] The viewpoint queue generation module is used to generate the viewpoint queue L through the viewpoint generation algorithm according to the input point cloud data. view ={p1,p2,...,p i ,...,p n}, where p i ∈R 3 ;

[0033] The cluster queue determination module is used to divide the above viewpoint queues into different clusters by viewpoint clustering to obtain cluster queues L clusters ={c1,c2,...,c j ,...,c m};

[0034] Cost value calculation module, used to calculate the current location cluster c in real time j1 The cost function value between the current clusters and other clusters, wherein the cost function includes a boundary heuristic function, a dynamic loss cost function and a distance cost function;

[0035] The exploration area determination module is used to iteratively select the cluster point with the smallest cost as the next expansion area based on the current exploration device information; repeat this process to guide the exploration device to explore the unknown area in real time.

[0036] On the other hand, the present invention also provides a computer electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned area exploration method based on point cloud clustering when executing the computer program.

[0037] On the other hand, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned area exploration method based on point cloud clustering are implemented.

[0038] Compared with the prior art, the present invention has the following advantages: the present invention optimizes the exploration order by clustering the viewpoints in the area through the point cloud viewpoints; then ensures the rationality of path planning and the comprehensiveness of exploration through a cost function integrating a boundary heuristic function, a dynamic loss cost function and a distance cost function, wherein the boundary heuristic function can dynamically adjust the exploration order according to the boundary information of the environment, avoid the problems of missed scans and incomplete coverage, and ensure comprehensive access to the target area; and the dynamic loss cost function can consider the kinematic characteristics of the exploration device (such as an autonomous vehicle, a robot or an unmanned aerial vehicle) in the path planning, optimize the moving path, reduce energy consumption and time cost, thereby improving the exploration efficiency in a complex dynamic environment, and the distance cost function can ensure the optimality of the path when exploring the area; finally, according to the current exploration device information, the cluster point with the smallest calculated cost value is selected as the next expansion area for exploration; repeating this process guides the exploration device to explore the unknown area in real time, thereby significantly improving the adaptability and exploration efficiency to complex environments, and is suitable for the three-dimensional space reconstruction task of robots, autonomous vehicles and unmanned aerial vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0040] Figure 1 It is a flow chart of the region exploration method based on point cloud clustering of the present invention;

[0041] Figure 2 It is a structural block diagram of the region exploration system based on point cloud clustering of the present invention;

[0042] Figure 3 Generate a viewpoint queue diagram for sampling;

[0043] Figure 4 A schematic diagram of viewpoint clustering during operation (different colors represent different clusters);

[0044] Figure 5 This is a schematic diagram of the effect of using the method of the present invention to perform regional exploration;

[0045] Figure 6 Schematic diagram of the greedy random search algorithm;

[0046] Figure 7 Block diagram of the region exploration device based on point cloud clustering. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] See also Figure 1 The embodiment of the present invention discloses a region exploration method based on point cloud clustering, comprising the following steps:

[0049] S1, based on the input point cloud data, generate the viewpoint queue L through the viewpoint generation algorithm view ={p1,p2,...,p i ,...,p n}, where viewpoint p i Belongs to the three-dimensional viewpoint;

[0050] S2, divide the above viewpoint queues into different clusters c through viewpoint clustering j, to obtain the clustering queue L clusters ={c1,c2,...,c j ,...,c m};

[0051] S3, real-time computing and exploring the current location cluster of the device j1 A cost function between the clusters and other current clusters, wherein the cost function includes a boundary heuristic function, a dynamic loss cost function and a distance cost function;

[0052] S4, based on the current exploration device information, select the cluster point with the smallest cost as the next expansion area; repeat this process to guide the exploration device to explore the unknown area in real time.

[0053] See also Figure 2 The embodiment of the present invention further discloses a region exploration system based on point cloud clustering, comprising:

[0054] The viewpoint queue generation module is used to generate the viewpoint queue L through the viewpoint generation algorithm according to the input point cloud data. view ={p1,p2,...,p i ,...,p n}, where viewpoint p i Belongs to the three-dimensional viewpoint;

[0055] The cluster queue determination module is used to divide the above viewpoint queues into different clusters by viewpoint clustering to obtain cluster queues L clusters ={c1,c2,...,c j ,...,c m};

[0056] Cost value calculation module, used to calculate the current location cluster c in real time j1 The cost function value between the current clusters and other clusters, wherein the cost function includes a boundary heuristic function, a dynamic loss cost function and a distance cost function;

[0057] The exploration area determination module is used to iteratively select the cluster point with the smallest cost as the next expansion area based on the current exploration device information; repeat this process to guide the exploration device to explore the unknown area in real time.

[0058] In this embodiment, the region exploration method based on point cloud clustering is to use the region exploration system based on point cloud clustering as the execution object of the steps. Specifically, step S1 is to use the viewpoint queue generation module as the execution object of the step, step S2 is to use the cluster queue determination module as the execution object of the step, step S3 is to use the cost value calculation module as the execution object of the step, and step S4 is to use the exploration region determination module as the execution object of the step.

[0059] Among them, in step S1, point cloud data in the exploration area is obtained through sensors such as laser radar (LiDAR), structured light, and stereo vision. These point cloud data can provide the precise position of the object surface in three-dimensional space. Therefore, in step S1, the point cloud data obtained by sensors such as laser radar (LiDAR), structured light, and stereo vision are input into the viewpoint generation algorithm, and the viewpoint generation algorithm generates a viewpoint queue; the current mainstream viewpoint generation algorithm, such as random sampling, boundary sampling, etc., can be used to generate global viewpoints. The viewpoint queue is also generated incrementally. Let the viewpoint queue be L view ={p1,p2,...,p i ,...,p n}, where p i ∈R 3 , i.e., viewpoint p i It belongs to the three-dimensional viewpoint. For a three-dimensional space with a fixed height, such as a tunnel or a fixed-height flight scene, the schematic diagram of the viewpoint generated by the mainstream method can be as follows Figure 3 shown.

[0060] Furthermore, since the point cloud data can provide the accurate position of the object surface in the three-dimensional space, it is necessary to ensure the validity of the input point cloud data to ensure the accurate position of the object surface in the three-dimensional space. Therefore, in order to ensure the validity of the point cloud data input into the viewpoint generation algorithm, before generating the viewpoint queue through the viewpoint generation algorithm according to the input point cloud data, the following steps are also included:

[0061] S0, incrementally constructs a point cloud map through lidar to detect the validity of the point cloud, and inputs the valid point cloud data into the viewpoint generation algorithm to generate a viewpoint queue.

[0062] In step S0, the validity of the point cloud is detected by incrementally constructing a point cloud map through the laser radar. When the point cloud data is normal and valid, it is input into the viewpoint generation algorithm, and the viewpoint queue L is obtained by sampling. view ={p1,p2,...,p i ,...,p n}.

[0063] Specifically, in step S2, the step of dividing the above viewpoint queues into different clusters by viewpoint clustering to obtain cluster queues includes:

[0064] S201, taking the current exploration device location as the cluster center cluster j , determine all unclustered viewpoints p i and cluster center cluster j Is the maximum cluster distance between them less than the preset cluster distance d? cluster ;

[0065] S202, through the unclustered viewpoint p i and cluster center cluster j Connection detection function Directlink(p i ,cluster j ) Determine the unclustered viewpoint p i and cluster center cluster j Whether there is a collision between them;

[0066] S203, when the unclustered viewpoint p i and cluster center cluster j The maximum cluster distance between them is less than the preset cluster distance d cluster And the unclustered viewpoint p i and cluster center cluster j When there is no collision between the cluster viewpoints p i Join the cluster c j Otherwise, p i Create a new cluster for the center;

[0067] S204, repeat the above process to obtain the cluster queue L clusters ={c1,c2,...,c j ,...,c m}, where each cluster c j Contains all viewpoints p that meet the clustering condition i .

[0068] In this embodiment, since the viewpoints generated in step S1 are disordered, this will bring great uncertainty to the exploration efficiency of the region. Therefore, this embodiment adopts a two-dimensional plane clustering method and only considers the plane coordinates of the viewpoints during clustering, which can reduce the calculation amount of one dimension.

[0069] Specifically, in step S201, the first clustering is centered on the current exploration device position, and all newly added and unclustered viewpoints p are checked. i , when it meets the following conditions, it will be added to the cluster cluster j middle:

[0070]

[0071] Directlink(p i ,cluster j ) is the connection detection function between the viewpoint and the cluster center, which is true if and only if there is no collision between the lines. In addition, Dis(p i ,cluster j) is p i With cluster j The distance calculation function between them satisfies the maximum clustering distance d cluster When the requirements are met.

[0072] In summary, in step S203, when the viewpoint p i and cluster center cluster j When the distance is met and there is no collision in the middle, the viewpoint p i Add to the cluster, otherwise use p i Create a new cluster for the center.

[0073] Repeating the above process can get a cluster queue: L clusters ={c1,c2,...,c j ,...,c m}, where each cluster c j Contains all viewpoints p that meet the clustering condition i The clustering effect can be seen in Figure 4 shown.

[0074] The cluster queue L determined by steps S201-S204 above clusters ={c1,c2,...,c j ,...,c m} is the cluster exploration queue of the exploration device. In order to obtain a reasonable cluster exploration order and improve the exploration efficiency between clusters, it is necessary to construct a conversion cost function f between clusters. total (c j1 ,c j2 ), the cost function f total (c j1 ,c j2 ) contains the boundary heuristic function f b (c j1 ,c j2 ), dynamic loss cost function f dy (c j1 ,c j2 ) and the distance cost function f d (c j1 ,c j2 ). The boundary heuristic function can dynamically adjust the exploration order according to the boundary information of the environment, avoid the problems of missed scans and incomplete coverage, and ensure comprehensive access to the target area; the dynamic loss cost function can consider the kinematic characteristics of the exploration device (such as autonomous equipment, robots or drones, etc.) in path planning, optimize the mobile path, reduce energy consumption and time cost, and thus improve the exploration efficiency in complex dynamic environments; and the distance cost function can ensure the optimality of the path when exploring the area.

[0075] In step S3, the current location cluster c of the device is explored by real-time calculation j1 The cost function value between the current clusters and the other clusters, and then select the cluster point with the smallest cost value from the calculated results as the next expansion area through step S4.

[0076] During the exploration process, steps S1-S4 are repeated in real time according to the position of the exploration device, which can effectively guide the exploration device to explore unknown areas, significantly improve the adaptability and exploration efficiency to complex environments, and is suitable for three-dimensional space reconstruction tasks of robots, autonomous vehicles and drones.

[0077] Specifically, in step S3, the cost function is f total (c j1 ,c j2 ) is:

[0078] f total (c j1 ,c j2 )=αf b (c j1 ,c j2 )+βf dy (c j1 ,c j2 )+γf d (c j1 ,c j2 ),

[0079] Among them, α, β, γ represent the weight ratio, f b (c j1 ,c j2 ) is the boundary heuristic function, f dy (c j1 ,c j2 ) is the dynamic loss cost function, f d (c j1 ,c j2 ) is the distance cost function.

[0080] Among them, α, β, and γ can be adjusted according to demand. Assume that the current location of the exploration device (such as drones, robots, autonomous vehicles, etc.) is cluster c j1 , update the clustering queue L in real time according to the point cloud clusters ={c1,c2,...,c j ,...,c m}, calculate the cluster c j1 The cost function f for each cluster that changes total (c j1 ,c j ), select the cluster point with the smallest cost value as the next expansion point area:

[0081]

[0082] Furthermore, the boundary heuristic function f b (c j1 ,c j2 ) value is determined by the following formula:

[0083] f b (c j1 ,c j2 )=GetBarrierDis(c j2 )-GetBarrierDis(c j1 )

[0084] GetBarrierDis(c j1 )=min(|c j1,x_center -x barrier |,|c j1,y_center -y barrier |)

[0085] where c j1,x_center 、c j1,y_center Represents cluster c j1 The average coordinate value of the viewpoint in two dimensions, x barrier ,y barrier Indicates the boundary value of the search area, GetBarrierDis(c j1 ) Calculate the distance plane c j1 The present embodiment introduces a boundary heuristic function to guide path selection during the exploration process. This function can dynamically adjust the exploration order according to the boundary information of the environment, avoid the problem of missed scans and incomplete coverage, and ensure full access to the target area. b (c j1 ,c j2 ) is smaller, indicating that it is closer to the boundary, guiding the robot to explore the boundary area first.

[0086] Furthermore, the dynamic loss cost function f dy (c j1 ,c j2 ) value is determined by the following formula:

[0087]

[0088] where c j1,center For cluster c j1 The average position of all viewpoints in j2,center For cluster c j2 The average position of all viewpoints in j1,θ For cluster cj1 The attitude angles of all viewpoints in j2,θ For cluster c j2 The pose angles of all viewpoints in .

[0089] This embodiment designs a dynamic loss cost function to take into account the kinematic characteristics of the robot or drone in path planning, optimize the moving path, reduce energy consumption and time cost, and thus improve exploration efficiency in a complex dynamic environment. dy (c j1 ,c j2 ) is used to measure the loss function of the robot dynamics and guide the drone to transition to the cluster with the smallest energy loss and conduct exploration.

[0090] Furthermore, the distance cost function f d (c j1 ,c j2 ) value is determined by the following formula:

[0091] f d (c j1 ,c j2 )=||c j1,center -c j2,center ||

[0092] where c j1,center For cluster c j1 The average position of all viewpoints in j2,center For cluster c j2 The average position of all viewpoints in the distance cost function f d (c j1 ,c j2 ) By using the Euclidean distance to measure the transition loss between clustering areas, it is possible to get as close as possible to the closer areas. This embodiment combines the distance cost function in the path planning process to ensure the optimality of the path when exploring the area. The embodiment of the present invention comprehensively considers the complexity of the environment and the requirements of the task, and further improves the rationality of the path planning.

[0093] The unknown area is explored by using the method of the present invention and the greedy random exploration algorithm. The schematic diagram of using the greedy random exploration algorithm to explore the unknown area is as follows: Figure 5 As shown, the schematic diagram of using the method of the present invention to explore unknown areas is as follows Figure 6 As shown in the figure, it can be seen that the regional exploration method based on point cloud clustering proposed in the present invention can effectively guide the robot to explore unknown areas, and less local omissions occur. Figure 6 It can be clearly seen that the method of the present invention can solve the problem of local exploration omissions and effectively improve the robot's exploration efficiency.

[0094] In summary, the present invention clusters viewpoints in the area into point clouds to optimize the order of exploration; then, a cost function integrating a boundary heuristic function, a dynamic loss cost function and a distance cost function is used to ensure the rationality of path planning and the comprehensiveness of exploration, wherein the boundary heuristic function can dynamically adjust the exploration order according to the boundary information of the environment to avoid the problems of missed scans and incomplete coverage, and ensure comprehensive access to the target area; the dynamic loss cost function can consider the kinematic characteristics of the robot or drone in path planning, optimize the moving path, reduce energy consumption and time cost, and thus improve the exploration efficiency in a complex dynamic environment, while the distance cost function can ensure the optimality of the path when exploring the area; finally, the cluster point with the smallest calculated cost value is selected as the next expansion area for exploration according to the current robot information; the process is repeated to guide the robot to explore unknown areas in real time, thereby significantly improving the adaptability and exploration efficiency to complex environments, and is suitable for three-dimensional space reconstruction tasks of robots, autonomous vehicles and drones.

[0095] On the other hand, Figure 7 As shown, an embodiment of the present invention further provides a computer electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned area exploration method based on point cloud clustering are implemented.

[0096] On the other hand, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned area exploration method based on point cloud clustering are implemented.

[0097] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A region exploration method based on point cloud clustering, characterized in that: The following steps are involved: According to the input point cloud data, the viewpoint queue L is generated through the viewpoint generation algorithm view ={p1,p2,...,p i ,...,p n }, where viewpoint p i Belongs to the three-dimensional viewpoint; The above viewpoint queues are divided into different clusters c by viewpoint clustering j , to obtain the clustering queue L clusters ={c1,c2,...,c j ,...,c m }; Real-time computing explores the current location cluster of the device j1 A cost function between the clusters and other current clusters, wherein the cost function includes a boundary heuristic function, a dynamic loss cost function and a distance cost function; According to the current exploration device information, select the cluster point with the smallest cost value as the next expansion area; This process is repeated to guide the exploration device to explore unknown areas in real time.

2. The region exploration method based on point cloud clustering according to claim 1, characterized in that: The method further includes the following steps before generating a viewpoint queue by a viewpoint generation algorithm according to the input point cloud data: The validity of the point cloud is detected by incrementally constructing a point cloud map through lidar, and the valid point cloud data is input into the viewpoint generation algorithm to generate a viewpoint queue.

3. The region exploration method based on point cloud clustering according to claim 1, characterized in that: The step of dividing the above viewpoint queues into different clusters by viewpoint clustering to obtain cluster queues includes: The current exploration device location is the cluster center cluster j , determine all unclustered viewpoints p i and cluster center cluster j Is the maximum cluster distance between them less than the preset cluster distance d? cluster ; Through the unclustered viewpoint p i and cluster center cluster j The connection detection function between them determines the unclustered viewpoints p i and cluster center cluster j Whether there is a collision between them; When the unclustered viewpoint p i and cluster center cluster j The maximum cluster distance between them is less than the preset cluster distance d cluster And the unclustered viewpoint p i and cluster center cluster j When there is no collision between the cluster viewpoints p i Join the cluster c j Otherwise, p i Create a new cluster for the center; Repeat the above process to obtain the cluster queue L clusters ={c1,c2,...,c j ,...,c m }, where each cluster c j Contains all viewpoints p that meet the clustering condition i .

4. The region exploration method based on point cloud clustering according to claim 1, characterized in that: The cost function is f total (c j1 ,c j2 ) is: f total (c j1 ,c j2 )=αf b (c j1 ,c j2 )+βf dy (c j1 ,c j2 )+γf d (c j1 ,c j2 ), Among them, α, β, γ represent the weight ratio, f b (c j1 ,c j2 ) is the boundary heuristic function, f dy (c j1 ,c j2 ) is the dynamic loss cost function, f d (c j1 ,c j2 ) is the distance cost function.

5. The region exploration method based on point cloud clustering according to claim 1, characterized in that: Boundary heuristic function f b (c j1 ,c j2 ) value is determined by the following formula: f b (c j1 ,c j2 )=GetBarrierDis(c j2 )-GetBarrierDis(c j1 ) GetBarrierDis(c j1 )=min(|c j1,x_center -x barrier |,|c j1,y_center -y barrier |) where c j1,x_center 、c j1,y_center Represents cluster c j1 The average coordinate value of the viewpoint in two dimensions, x barrier ,y barrier Indicates the boundary value of the search area, GetBarrierDis(c j1 ) Calculate the distance plane c j1 Distance from the border.

6. The region exploration method based on point cloud clustering according to claim 1, characterized in that: Dynamic loss cost function f dy (c j1 ,c j2 ) value is determined by the following formula: where c j1,center For cluster c j1 The average position of all viewpoints in j2,center For cluster c j2 The average position of all viewpoints in j1,θ For cluster c j1 The attitude angles of all viewpoints in j2,θ For cluster c j2 The pose angles of all viewpoints in .

7. The region exploration method based on point cloud clustering according to claim 1, characterized in that: Distance cost function f d (c j1 ,c j2 ) value is determined by the following formula: f d (c j1 ,c j2 )=||c j1,center -c j2,center || where c j1,center For cluster c j1 The average position of all viewpoints in j2,center For cluster c j2 The average position of all viewpoints in .

8. A region exploration system based on point cloud clustering, characterized in that: include: The viewpoint queue generation module is used to generate the viewpoint queue L through the viewpoint generation algorithm according to the input point cloud data. view ={p1,p2,...,p i ,...,p n }, where viewpoint p i Belongs to the three-dimensional viewpoint; The cluster queue determination module is used to divide the above viewpoint queues into different clusters by viewpoint clustering to obtain cluster queues L clusters ={c1,c2,...,c j ,...,c m }; Cost value calculation module, used to calculate the current location cluster c in real time j1 The cost function value between the cluster and other current clusters, wherein the cost function includes a boundary heuristic function, a dynamic loss cost function and a distance cost function; An exploration area determination module is used to iteratively select the cluster point with the smallest cost value as the next expansion area according to the current exploration device information; This process is repeated to guide the exploration device to explore unknown areas in real time.

9. A computer electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the area exploration method based on point cloud clustering according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the region exploration method based on point cloud clustering according to any one of claims 1 to 7 are implemented.

Images