Cutting Method, Device and System for Point Cloud Model

Abstract

The present invention provides a method, apparatus and system for cutting a point cloud model, which includes the following steps: S1, selecting a point cloud structure including a target object from a point cloud model by using a two-dimensional first cutting window; S2, adjusting the depth of the first cutting window, and the length, width and depth of the first cutting window form a three-dimensional second cutting window, and the target object is located in the second cutting window; S3, identifying and marking all point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, wherein the target object is located in one of the third cutting windows; S4, calculating the volume ratio of the point cloud structure in each third cutting window relative to the second cutting window, and selecting the third cutting window with the largest volume ratio. The cutting mechanism of the point cloud model provided by the present invention can consider the depth information of the 3D point cloud model that was originally omitted, and the present invention can automatically screen out the target object and send it to the customer.

Description

Technical Field

[0001] The present invention relates to the field of modeling, and in particular, to a method, apparatus, and system for cutting a point cloud model. Background Art

[0002] Nowadays, the application of RGB-D cameras has received increasing attention. Compared with RGB cameras, RGB-D cameras can provide point cloud images with depth information, which means that many scenes in the real world can be captured by RGB-D cameras, such as digital twins in autonomous factories and environment awareness in autonomous robots. Therefore, the point cloud images in RGB-D cameras have received intensive research.

[0003] For end users who can obtain different types of RGB-D cameras, RGB-D cameras can help them easily obtain color and depth information. However, for end users, RGB-D cameras only have limited techniques and tool geometries to provide autonomous 3D data for pruning and editing.

[0004] Point cloud cutting, as a main method for processing point cloud images, is very important in the field of computer vision because it can select targets for many secondary tasks from point cloud images. Among them, secondary tasks include point cloud registration, point cloud localization, and robot grasping, etc. However, the sensors themselves can introduce a large amount of noise. In addition, the unprocessed point cloud images include a lot of background or irrelevant point clouds of other targets. Such irrelevant point clouds will greatly hinder the cutting of 3D models, so a practical 3D point cloud cutting tool must be used. Traditional point cloud cutting tools do not allow users to select a specific depth from a certain viewpoint of the point cloud image, which means that the point clouds desired by users cannot always be selected. Moreover, traditional point cloud cutting tools rely on user interfaces without semantic functions, which results in users not being able to automatically select the targets they want.

[0005] Until now, methods for semi-automatic or automatic pruning and extraction of 3D point cloud data to obtain target geometric features are still major problems to be solved for 3D reconstruction and 3D robot perception. The prior art also provides several manual methods or tools, such as CloudCompare3D, which can assist users in selecting and pruning 3D point cloud data. However, such software does not support semi-automatic or automatic methods. In addition, when users select or prune 3D point cloud data, the CloudCompare3D mechanism lacks depth information.

[0006] The prior art also provides some 3D point cloud cutting mechanisms, which provide a system including an optical camera data processing unit, capable of obtaining a 3D point cloud scene including a target object. Using a switching input device, a user can input a seed, where the seed indicates the position of the target object. Finally, the segmentation method generates a segmented point cloud corresponding to the target object by pruning the 3D point cloud based on the position reference input by the user. Summary of the Invention

[0007] A first aspect of the present invention provides a method for cutting a point cloud model, which includes the following steps: S1, selecting a point cloud structure including a target object from a point cloud model using a two-dimensional first cutting window, where the first cutting window has a length and a width; S2, adjusting the depth of the first cutting window, and the length, width, and depth of the first cutting window form a three-dimensional second cutting window, and the target object is located in the second cutting window; S3, identifying and marking all point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, where the target object is located in one of the third cutting windows; S4, calculating the volume ratio of the point cloud structure in each third cutting window relative to the second cutting window, selecting the third cutting window with the largest volume ratio, and determining that the point cloud structure in the third cutting window is the target object, where the third cutting window is smaller than the second cutting window, and the second cutting window is smaller than the first cutting window.

[0008] Further, step S3 further includes the following steps: S31, calculating the number k of the third cutting windows; S32, randomly selecting k points from all point cloud structures in the second cutting window as centroids, then calculating the distances from other points in all point cloud structures to the centroids with the seed centroids as the clustering centers, and assigning other points in all point cloud structures to the nearest centroid to form a cluster, and iteratively executing steps S31 and S32 until the positions of the k centroids no longer change; S33, identifying and marking all point cloud structures in the second cutting window to form k three-dimensional third cutting windows, where the k third cutting windows include k clusters.

[0009] Further, step S3 further includes the following steps: training a sample of the target object using the dataset of the target object, comparing the sample of the target object with the target object, and identifying and marking all point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, where the target object is located in one of the third cutting windows.

[0010] The second aspect of the present invention provides a cutting system for a point cloud model, including: a processor; and a memory coupled to the processor, the memory having instructions stored therein, the instructions, when executed by the processor, causing the electronic device to perform actions, the actions including: S1, selecting a point cloud structure including a target object from a point cloud model by using a two-dimensional first cutting window, the first cutting window having a length and a width; S2, adjusting the depth of the first cutting window, the length, width and depth of the first cutting window forming a three-dimensional second cutting window, the target object being located in the second cutting window; S3, identifying and marking all point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, wherein the target object is located in one of the third cutting windows; S4, calculating the volume ratio of the point cloud structure in each third cutting window relative to the second cutting window, selecting the third cutting window with the largest volume ratio and determining that the point cloud structure in the third cutting window is the target object, wherein the third cutting window is smaller than the second cutting window, and the second cutting window is smaller than the first cutting window.

[0011] Further, the action S3 further includes: S31, calculating the number k of the third cutting windows; S32, randomly selecting k points from all point cloud structures in the second cutting window as centroids, then calculating the distances from other points in all point cloud structures to the centroids with the seed centroids as the clustering centers, and assigning other points in all point cloud structures to the nearest centroid to form a cluster, iteratively executing steps S31 and S32 until the positions of the k centroids no longer change; S33, identifying and marking all point cloud structures in the second cutting window to form k three-dimensional third cutting windows, wherein the k third cutting windows include k clusters.

[0012] Further, the action S3 further includes: training a sample of the target object by using the data set of the target object, comparing the sample of the target object with the target object, so as to identify and mark all point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, wherein the target object is located in one of the third cutting windows.

[0013] The third aspect of the present invention provides a cutting device for a point cloud model, which includes: a first cutting device that selects a point cloud structure including a target object from a point cloud model by using a two-dimensional first cutting window, and the first cutting window has a length and a width; a depth adjustment device that adjusts the depth of the first cutting window, and the length, width, and depth of the first cutting window form a three-dimensional second cutting window, and the target object is located in the second cutting window; a second cutting device that identifies and marks all the point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, where the target object is located in one of the third cutting windows; a calculation device that calculates the volume ratio of the point cloud structure in each third cutting window to the second cutting window, selects the third cutting window with the largest volume ratio, and determines that the point cloud structure in the third cutting window is the target object, where the third cutting window is smaller than the second cutting window, and the second cutting window is smaller than the first cutting window.

[0014] Further, the second cutting device is further configured to calculate the number k of the third cutting windows, randomly select k points from all the point cloud structures in the second cutting window as centroids, then calculate the distances from the other points in all the point cloud structures to the centroids with the seed centroids as the clustering centers, and assign the other points in all the point cloud structures to the nearest centroid to form a cluster until the positions of the k centroids no longer change, and identify and mark all the point cloud structures in the second cutting window to form k three-dimensional third cutting windows, where the k third cutting windows include k clusters.

[0015] Further, the second cutting device is further configured to train a sample of the target object by using the data set of the target object, and compare the sample of the target object with the target object to identify and mark all the point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, where the target object is located in one of the third cutting windows.

[0016] The fourth aspect of the present invention provides a computer program product, which is tangibly stored on a computer-readable medium and includes computer-executable instructions, and the computer-executable instructions, when executed, cause at least one processor to execute the method according to the first aspect of the present invention.

[0017] The fifth aspect of the present invention provides a computer-readable medium, on which computer-executable instructions are stored, and the computer-executable instructions, when executed, cause at least one processor to execute the method according to the first aspect of the present invention.

[0018] The cutting mechanism of the point cloud model provided by the present invention can consider the depth information of the 3D point cloud model that was originally omitted, and the present invention can automatically screen out the target object and send it to the customer. In addition, the present invention uses methods such as clustering and deep learning to screen the target object. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is an architecture diagram of a cutting system for a point cloud model according to a specific embodiment of the present invention;

[0020] Figure 2 is a schematic diagram of a point cloud model and a first cutting window of the cutting mechanism of the point cloud model according to a specific embodiment of the present invention;

[0021] Figure 3 is a schematic diagram of a target object and a second cutting window of the point cloud model of the cutting mechanism of the point cloud model according to a specific embodiment of the present invention;

[0022] Figure 4 is a schematic diagram of a target object and a third cutting window of the point cloud model of the cutting mechanism of the point cloud model according to a specific embodiment of the present invention;

[0023] Figure 5 is a schematic diagram of the clustering method of the cutting mechanism of the point cloud model according to a specific embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] The following will describe the specific embodiments of the present invention with reference to the accompanying drawings.

[0025] The present invention provides a cutting mechanism for a point cloud model, which uses a three-dimensional cutting window to accurately lock the target object in the point cloud model and uses the volume ratio to select the target object therein.

[0026] As Figure 1 shown, the cutting system of the point cloud model includes a software module and a hardware device. The hardware device includes a screen S and a computing device D. The screen S has a hardware interface for the computing device D, such as an hdmi or VGA port, which has the ability to display the graphic data of the computing device D and display the data to the customer C. The computing device D has hardware interfaces with the screen S, the mouse M, and the keyboard K, and has the computing ability to download the point cloud model or the point cloud structure. The mouse M and the keyboard K are input devices of the customer C, and the computing device D can display data to the customer C through the screen S.

[0027] The software module includes a first cutting device 110, a downloading device 120, a depth adjusting device 130, a generating device 140, a second cutting device 150, a calculating device 160, a recommending device 170, and 180. Among them, the downloading device 120 is used to download a large amount of data of a point cloud model 200 and display the point cloud model 200 on the screen S. The first cutting device 110 selects a point cloud structure including the target object from the point cloud model by using a two-dimensional first cutting window. The depth adjusting device 130 adjusts the depth of the first cutting window to form a three-dimensional second cutting window. The generating device 140 receives the configurations and parameters of the downloading device 120 and the depth adjusting device 130 and generates a second cutting window serving as a limit frame based on the user input. The second cutting device 150 identifies and marks all the point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, where the target object is located in one of the third cutting windows. The calculating device 160 calculates the volume ratio of the point cloud structure in each third cutting window relative to the second cutting window, selects the third cutting window with the largest volume ratio, and determines that the point cloud structure in the third cutting window is the target object.

[0028] The first aspect of the present invention provides a method for cutting a point cloud model, which includes the following steps:

[0029] First, step S1 is executed. The first cutting device 110 selects a point cloud structure including the target object from a point cloud model by using a two-dimensional first cutting window, and the first cutting window has a length and a width.

[0030] Among them, as Figure 2 shown, the downloading device 120 is used to download a large amount of data of a point cloud model 200 and display the point cloud model 200 on the screen S. In this embodiment, the point cloud model 200 includes a first cylinder 210, a second cylinder 220, a first cube 230, a second cube 240, and other redundant point cloud structures (not shown). Among them, the first cylinder 210, the second cylinder 220, the first cube 230, and the second cube 240 are all point cloud structures. The target object among them is the first cylinder 210.

[0031] It should be noted that Figure 2 the shown first cylinder 210, second cylinder 220, first cube 230, and second cube 240 are all point cloud structures, Figure 3 and Figure 4 all the objects shown in Figure 3 are also point cloud structures. For the sake of convenience and simplicity of description, the point cloud structures are omitted. That is, Figure 4 a part of the first cylinder 210 and the second cylinder 220 in Figure 4The first cylinder 210, the first redundant point cloud structure 250, and the second redundant point cloud structure 260 in it are also all point cloud structures.

[0032] Specifically, the first cutting device 110 obtains the position of the mouse M relative to the screen S input by the user C through the keyboard K and the mouse M to generate a rectangular first cutting window W1. Among them, the first cutting window W1 is two-dimensional and only has a length l and a width h, without a depth. As Figure 2 shown, the first cylinder 210 as the target object is accommodated in the first cutting window W1. In addition, there are also some redundant point cloud structures (not shown) in the first cutting window W1.

[0033] Then, step S2 is executed, and the depth adjustment device 130 adjusts the depth of the first cutting window. The length, width, and depth of the first cutting window form a three-dimensional second cutting window, and the target object is located in the second cutting window.

[0034] Among them, the depth adjustment device 130 will automatically generate a slider (not shown) for the user. The user C slides the slider on the screen S through the mouse M to input the desired depth. Among them, the slider can display two end values of the minimum depth and the maximum depth for the user to select. As Figure 3 shown, the depth represented by the slider before the user input is d', and after the user inputs the desired depth, the depth is adjusted from d' to d. At this time, the length l, width h, and depth d of the first cutting window form a three-dimensional second cutting window W2, and the first cylinder 210 as the target object is located in the second cutting window W2. At this time, a part of the second cylinder 220 was originally in the cutting window, and at this time, through the adjustment of the depth, a part of the second cylinder 220 is no longer accommodated in the second cutting window W2.

[0035] In addition, the user C can also switch the view and angle of the point cloud model 200 displayed on the screen S through the mouse M. Comparing Figure 3 and Figure 4 , the view angle and angle of the point cloud model 200 are different, and by adjusting the view angle and angle, the depth of the second cutting window W2 can be better adjusted.

[0036] After the user C selects a satisfactory depth, the generation device 140 receives the configurations and parameters of the download device 120 and the depth adjustment device 130, and generates the second cutting window W2 serving as a limit frame based on the user's input.

[0037] Then step S3 is executed, and the second cutting device 150 identifies and marks all the point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, where the target object is located in one of the third cutting windows.

[0038] Step S3 can be implemented in various ways, such as clustering methods, deep learning methods, and supervoxel clustering methods.

[0039] Among them, according to the clustering method, step S3 further includes sub-step S31, sub-step S32, and sub-step S33.

[0040] In sub-step S31, calculate the number k of the third cutting windows. As Figure 4 shown, in the second cutting window W2, only a rough selection is made, which contains the first cylinder 210 of the target object and other redundant point cloud structures. Among them, the redundant point cloud structures include the first redundant point cloud structure 250 and the second redundant point cloud structure 260. The point cloud structures in each second window W2 are all accommodated by the third cutting window. Specifically, the first cylinder 210 is in the third cutting window W 31 in, the first redundant point cloud structure 250 is in the third cutting window W 32 in, the second redundant point cloud structure 260 is in the third cutting window W 33 in, so k = 3.

[0041] In sub-step S32, randomly select k points from all the point cloud structures in the second cutting window as centroids, and then calculate the distances from the other points in all the point cloud structures to the centroids with the seed centroids as the clustering centers, and assign the other points in all the point cloud structures to the centroid with the closest distance to form a cluster.

[0042] Iteratively execute steps S31 and S32 until the positions of the k centroids no longer change.

[0043] Finally, execute step S33 to identify and label all the point cloud structures in the second cutting window to form k three-dimensional third cutting windows, where the k third cutting windows include k clusters.

[0044] Specifically, Figure 5 illustrates the principle of the clustering method. Figure 5 includes multiple point cloud structures, where three points are selected as centroids, namely the first centroid z1, the second centroid z2, and the third centroid z3. Then, with the first centroid z1, the second centroid z2, and the third centroid z3 as the clustering centers, calculate Figure 5 the distances from the other points in all the point cloud structures to the first centroid z1, the second centroid z2, and the third centroid z3 respectively. Taking the third centroid z3, the distances between the third centroid z3 and all the points are d1, d2... d n , and then assign the points in all the point cloud structures with the closest distance to the third centroid z3 to form a cluster with the third centroid z3. And so on, the first centroid z1 and the second centroid z2 also respectively form a cluster. At this time, we have calculated three clusters.

[0045] In this embodiment, randomly select Figure 4 Any 3 points from all the point cloud structures in as the centroids, and according to the above clustering principle, three clusters can be separated, namely, the first cluster of the first cylinder 210 as the target object, the second cluster of the first redundant point cloud structure 250, and the third cluster of the second redundant point cloud structure 260.

[0046] According to the deep learning method, the step S3 further includes the following steps: using the data set of the target object to train the samples of the target object, using the samples of the target object to compare with the target object, so as to identify and label all the point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, wherein the target object is located in one of the third cutting windows.

[0047] According to the supervoxel clustering method, calculate the number of the third cutting windows, and then randomly select k points from all the point cloud structures in the second cutting window as seed voxels, and use LCCP to mark the concave-convex relationship of different surfaces, so as to divide the points without intersecting regions in the second cutting window into different supervoxel clusters, thereby identifying and labeling all the point cloud structures in the second cutting window to form k three-dimensional third cutting windows, where the k third cutting windows include k clusters.

[0048] Specifically, supervoxel clustering first selects several seed voxels, and then performs region growing to become large point cloud clusters. The points within each point cloud cluster are similar. Because supervoxel clustering is over-segmentation, that is, it is possible that adjacent objects are clustered together, and supervoxel clustering will segment them into one object. Then LCCP is used to mark the concave-convex relationship of different surfaces, and then the region growing algorithm is used to cluster small regions into large objects. This small region growing algorithm is restricted by concavity and convexity, that is: only allowing the region to grow across convex edges. Since there is no intersecting region between the first cylinder 210 and the first redundant point cloud structure 250 in the figure, separate supervoxel clustering can segment them into different point clouds.

[0049] Specifically, in this embodiment, a plurality of data related to the first cylinder 210 (such as the CAD model of the first cylinder 210) are used as a data set to train the sample of the first cylinder 210. This sample is used to compare with the first cylinder 210 to identify and label all the point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows. Among them, the first cylinder 210 is located in one of the third cutting windows. Object recognition is based on the user knowing what the target object to be found is. The target object has boundaries and is very different from other impurities. Based on this, binary classification is performed to distinguish the target object from the impurities. A large amount of data is used to train the features of the object. Each sample has a corresponding label. During the training process, the results are continuously compared with the labels to reduce errors.

[0050] Finally, step S4 is executed. The computing device 160 calculates the volume ratio of the point cloud structure in each third cutting window relative to the second cutting window, selects the third cutting window with the largest volume ratio, and determines that the point cloud structure in the third cutting window is the target object.

[0051] The computing device 160 calculates the volumes of the first cylinder 21, the first redundant point cloud structure 250, and the second redundant point cloud structure 260 respectively. Assume that the volume of the first cylinder 210 in the third cutting window W31 is V1, and the first redundant point cloud structure 250 in the third cutting window W 32 has a volume of V2, and the second redundant point cloud structure 260 in the third cutting window W 33 has a volume of V3. Among them, the volume of the second cutting window W2 is V. Therefore, the volume ratio of the first cylinder 210 is

[0052] The volume ratios of the first redundant point cloud structure 250 and the second redundant point cloud structure 260 are respectively

[0053] and

[0054] When the following conditions are met:

[0055] and

[0056] Then,

[0057] it is determined that the third cutting window with the largest volume ratio is the target object, that is, the first cylinder 210 is the target object.

[0058] Among them, the third cutting window is smaller than the second cutting window, and the second cutting window is smaller than the first cutting window. Therefore, the target object 180 recommended by the recommendation device 170 is recommended to customer C, and the target object is the first cylinder 210.

[0059] In a second aspect of the present invention, a cutting system for a point cloud model is provided, including: a processor; and a memory coupled to the processor, the memory having instructions stored therein, the instructions, when executed by the processor, causing the electronic device to perform operations, the operations including: S1, selecting a point cloud structure including a target object from a point cloud model by using a two-dimensional first cutting window, the first cutting window having a length and a width; S2, adjusting the depth of the first cutting window, the length, width, and depth of the first cutting window constituting a three-dimensional second cutting window, the target object being located in the second cutting window; S3, identifying and marking all point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, wherein the target object is located in one of the third cutting windows; S4, calculating the volume ratio of the point cloud structure in each third cutting window to the second cutting window, selecting the third cutting window with the largest volume ratio, and determining that the point cloud structure in the third cutting window is the target object, wherein the third cutting window is smaller than the second cutting window, and the second cutting window is smaller than the first cutting window.

[0060] Further, the operation S3 further includes: S31, calculating the number k of the third cutting windows; S32, randomly selecting k points from all point cloud structures in the second cutting window as centroids, then calculating the distances from other points in all point cloud structures to the centroids with the seed centroids as the clustering centers, and assigning other points in all point cloud structures to the nearest centroid to form a cluster, and iteratively executing steps S31 and S32 until the positions of the k centroids no longer change; S33, identifying and marking all point cloud structures in the second cutting window to form k three-dimensional third cutting windows, where the k third cutting windows include k clusters.

[0061] Further, the operation S3 further includes: training a sample of the target object by using the data set of the target object, and comparing the sample of the target object with the target object to identify and mark all point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, wherein the target object is located in one of the third cutting windows.

[0062] The third aspect of the present invention provides a cutting device for a point cloud model, which includes: a first cutting device that selects a point cloud structure including a target object from a point cloud model by using a two-dimensional first cutting window, and the first cutting window has a length and a width; a depth adjustment device that adjusts the depth of the first cutting window, and the length, width, and depth of the first cutting window form a three-dimensional second cutting window, and the target object is located in the second cutting window; a second cutting device that identifies and marks all the point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, wherein the target object is located in one of the third cutting windows; a calculation device that calculates the volume ratio of the point cloud structure in each third cutting window to the second cutting window, selects the third cutting window with the largest volume ratio, and determines that the point cloud structure in the third cutting window is the target object, wherein the third cutting window is smaller than the second cutting window, and the second cutting window is smaller than the first cutting window.

[0063] Further, the second cutting device is further configured to calculate the number k of the third cutting windows, randomly select k points from all the point cloud structures in the second cutting window as centroids, then calculate the distances from the other points in all the point cloud structures to the centroids with the seed centroids as the clustering centers, and assign the other points in all the point cloud structures to the centroid with the closest distance to form a cluster until the positions of the k centroids no longer change, and identify and mark all the point cloud structures in the second cutting window to form k three-dimensional third cutting windows, where the k third cutting windows include k clusters.

[0064] Further, the second cutting device is further configured to train a sample of the target object by using the data set of the target object, compare the target object with the sample of the target object, and identify and mark all the point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, wherein the target object is located in one of the third cutting windows.

[0065] The fourth aspect of the present invention provides a computer program product, which is tangibly stored on a computer-readable medium and includes computer-executable instructions, and the computer-executable instructions, when executed, cause at least one processor to execute the method according to the first aspect of the present invention.

[0066] The fifth aspect of the present invention provides a computer-readable medium, on which computer-executable instructions are stored, and the computer-executable instructions, when executed, cause at least one processor to execute the method according to the first aspect of the present invention.

[0067] The cutting mechanism of the point cloud model provided by the present invention can consider the depth information of the 3D point cloud model that was originally omitted, and the present invention can automatically screen out the target object and send it to the customer. In addition, the present invention uses clustering methods and deep learning methods to screen the target object.

[0068] Although the content of the present invention has been described in detail through the above preferred embodiments, it should be recognized that the above description should not be considered as a limitation of the present invention. After those skilled in the art have read the above content, various modifications and substitutions to the present invention will be obvious. Therefore, the protection scope of the present invention should be defined by the appended claims. In addition, any reference signs in the claims should not be construed as limiting the claims concerned; the word "comprising" does not exclude other devices or steps not listed in the claims or the specification; words such as "first", "second", etc. are only used to denote names and do not denote any particular order.

Claims

1. A method for cutting a point cloud model, wherein, Including the following steps: S1. Select a point cloud structure including the target object from a point cloud model by using a two-dimensional first cutting window, where the first cutting window has a length and a width; S2. Adjust the depth of the first cutting window, and the length, width, and depth of the first cutting window form a three-dimensional second cutting window, and the target object is located in the second cutting window; S3. Identify and label all point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, where the target object is located in one of the third cutting windows; S4. Calculate the volume ratio of the point cloud structure in each third cutting window relative to the second cutting window, select the third cutting window with the largest volume ratio, and determine that the point cloud structure in the third cutting window is the target object, where the third cutting window is smaller than the second cutting window, and the second cutting window is smaller than the first cutting window; Step S3 further includes the following steps: S31. Calculate the number k of the third cutting windows; S32. Randomly select k points from all the point cloud structures in the second cutting window as centroids, then calculate the distances from the other points in all the point cloud structures to the centroids with the seed centroids as the clustering centers, and assign the other points in all the point cloud structures to the nearest centroid to form a cluster, Iteratively execute steps S31 and S32 until the positions of the k centroids no longer change, S33. Identify and label all point cloud structures in the second cutting window to form k three-dimensional third cutting windows, where the k third cutting windows include k clusters.

2. The cutting method of the point cloud model according to claim 1, characterized in that, Step S3 further includes the following steps: Use the data set of the target object to train the samples of the target object, use the samples of the target object to compare with the target object, so as to identify and label all point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, where the target object is located in one of the third cutting windows.

3. A cutting system for a point cloud model, including: A processor; And A memory coupled to the processor, where the memory has instructions stored therein, and the instructions, when executed by the processor, cause the electronic device to perform actions, and the actions include: S1. Select a point cloud structure including the target object from a point cloud model by using a two-dimensional first cutting window, where the first cutting window has a length and a width; S2. Adjust the depth of the first cutting window, and the length, width, and depth of the first cutting window form a three-dimensional second cutting window, and the target object is located in the second cutting window; S3. Identify and label all point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, where the target object is located in one of the third cutting windows; S4. Calculate the volume ratio of the point cloud structure in each third cutting window relative to the second cutting window, select the third cutting window with the largest volume ratio, and determine that the point cloud structure in the third cutting window is the target object, Wherein the third cutting window is smaller than the second cutting window, and the second cutting window is smaller than the first cutting window; The operation S3 further includes: S31, calculating the number k of the third cutting windows; S32, randomly selecting k points from all the point cloud structures in the second cutting window as the centroids, then calculating the distances from the other points in all the point cloud structures to the centroids with the seed centroids as the clustering centers, and assigning the other points in all the point cloud structures to the centroid with the closest distance to form a cluster; Iteratively execute steps S31 and S32 until the positions of the k centroids no longer change; S33, identifying and marking all the point cloud structures in the second cutting window to form k three-dimensional third cutting windows, where the k third cutting windows include k clusters.

4. The cutting system for the point cloud model according to claim 3, characterized in that, The operation S3 further includes: Using the data set of the target object to train the samples of the target object, and using the samples of the target object to compare with the target object to identify and mark all the point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, wherein the target object is located in one of the third cutting windows.

5. Cutting device for point cloud model, wherein, Including: A first cutting device, which selects the point cloud structure including the target object from a point cloud model by using a two-dimensional first cutting window, and the first cutting window has a length and a width; A depth adjusting device, which adjusts the depth of the first cutting window, and the length, width and depth of the first cutting window form a three-dimensional second cutting window, and the target object is located in the second cutting window; A second cutting device, which identifies and marks all the point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, wherein the target object is located in one of the third cutting windows; A calculating device, which calculates the volume ratio of the point cloud structure in each third cutting window to the second cutting window, selects the third cutting window with the largest volume ratio and determines that the point cloud structure in the third cutting window is the target object; Wherein the third cutting window is smaller than the second cutting window, and the second cutting window is smaller than the first cutting window; The second cutting device is further configured to calculate the number k of the third cutting windows, randomly select k points from all the point cloud structures in the second cutting window as the centroids, then calculate the distances from the other points in all the point cloud structures to the centroids with the seed centroids as the clustering centers, and assign the other points in all the point cloud structures to the centroid with the closest distance to form a cluster until the positions of the k centroids no longer change, and identify and mark all the point cloud structures in the second cutting window to form k three-dimensional third cutting windows, where the k third cutting windows include k clusters.

6. The cutting device for the point cloud model according to claim 5, characterized in that, The second cutting device is further configured to use the data set of the target object to train the samples of the target object, and use the samples of the target object to compare with the target object to identify and mark all the point cloud structures in the second cutting window to form a plurality of three-dimensional third cutting windows, wherein the target object is located in one of the third cutting windows.

7. A computer program product tangibly stored on a computer-readable medium and comprising computer-executable instructions that, when executed, cause at least one processor to perform the method according to claim 1 or 2.

8. A computer-readable medium having stored thereon computer-executable instructions that, when executed, cause at least one processor to perform the method according to claim 1 or 2.

Images