Remaining material plate identification method and device

Abstract

The invention provides an excess material plate identification method and device, and the method comprises the steps: taking a first point cloud obtained after the excess material plate on a machine tool is emptied and a corresponding first feature as a background, and obtaining a second point cloud obtained after the excess material plate is placed on the machine tool and a corresponding second feature; and performing feature matching on the first feature and the second feature to determine the position of the excess material plate on the machine tool. Wherein the first point cloud is used for representing the surface of the sword grating on the machine tool, and the second point cloud is used for representing the upper surface of the excess material plate and the surface, not covered by the excess material plate, of the sword grating, so that the complex slag background can be effectively filtered out by comparing the first feature corresponding to the first point cloud with the second feature corresponding to the second point cloud; the point cloud quality of the segmented excess material plate is improved, so that the difficulty of subsequent point cloud segmentation and contour extraction tasks is reduced, and finally the purpose of improving the algorithm performance and the working efficiency is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of integrated computer graphics, and in particular to a scrap plate recognition method and device. BACKGROUND

[0002] The flat area for carrying scrap plates on a laser cutting machine is generally a plurality of sawtooth-shaped metals, which are also known as laser cutting sword gratings or sword gratings in the market. In the prior art, scrap plates are recognized by distinguishing scrap plates from sword gratings for carrying scrap plates.

[0003] However, in the laser cutting scene, especially in the high-power thick plate cutting scene, a part of the molten metal will accumulate on the rack and re-solidify to form an irregular metal shape, which is called slag. Since the area where the slag is located is easily misidentified as part of the scrap plate, it is difficult to accurately identify the scrap plate in actual application. SUMMARY

[0004] The present application provides a scrap plate recognition method and device, which can accurately identify the position of the scrap plate.

[0005] According to a first aspect of the present application, a scrap plate recognition method is provided, comprising: After emptying the scrap plate on the machine, the surface of the sword grating after the scrap plate is emptied is scanned to obtain a first point cloud; The first point cloud is feature extracted to obtain a first feature; After placing a scrap plate on the machine, the upper surface of the scrap plate and the surface of the sword grating not covered by the scrap plate are scanned to obtain a second point cloud; The second point cloud is feature extracted to obtain a second feature; The first feature and the second feature are feature matched to determine the position of the scrap plate on the machine.

[0006] Optionally, the method for feature extracting the first point cloud to obtain the first feature comprises: obtaining the first feature by the FPFH Estimation method according to the first point cloud; And / or, the method for feature extracting the second point cloud to obtain the second feature comprises: obtaining the second feature by the FPFH Estimation method according to the second point cloud.

[0007] Optionally, the first point cloud includes a plurality of first sub-points, and the method of obtaining the first feature according to the first point cloud by the FPFH Estimation method includes: obtaining a normal of each first sub-point; constructing a first KD tree according to the first point cloud, the first KD tree being used for indexing the first sub-points in the first point cloud; and performing a nearest neighbor search on the first point cloud according to the first KD tree and the normal of each first sub-point to construct the first feature. And / or, the second point cloud includes a plurality of second sub-points, and the method of obtaining the second feature according to the second point cloud by the FPFH Estimation method includes: obtaining a normal of each second sub-point; constructing a second KD tree according to the second point cloud, the second KD tree being used for indexing the second sub-points in the second point cloud; and performing a nearest neighbor search on the second point cloud according to the second KD tree and the normal of each second sub-point to construct the second feature.

[0008] Optionally, the first point cloud includes a plurality of first sub-points, and before the first feature is obtained according to the first point cloud by the FPFH Estimation method, the method further includes: performing voxel filtering on the first point cloud to obtain a first sub-point cloud, the number of first sub-points in the first sub-point cloud being less than the number of first sub-points in the first point cloud. And / or, the second point cloud includes a plurality of second sub-points, and before the second feature is obtained according to the second point cloud by the FPFH Estimation method, the method further includes: performing voxel filtering on the second point cloud to obtain a second sub-point cloud, the number of second sub-points in the second sub-point cloud being less than the number of second sub-points in the second point cloud.

[0009] Optionally, after the first point cloud is voxel filtered to obtain a first sub-point cloud, the method further includes: extracting a plurality of first sub-points in the first sub-point cloud as a plurality of first key points by the ISSKeypoint 3D method, the number of first key points being less than the number of first sub-points in the first sub-point cloud. And / or, after the second point cloud is voxel filtered to obtain a second sub-point cloud, the method further includes: extracting a plurality of second sub-points in the second sub-point cloud as a plurality of second key points by the ISSKeypoint 3D method, the number of second key points being less than the number of second sub-points in the second sub-point cloud.

[0010] Optionally, the second point cloud includes a plurality of second sub-points, the first features include a plurality of first descriptors, the second features include a plurality of second descriptors, and the method of feature matching the first features and the second features to determine the position of the excess sheet on the machine tool includes: determining a plurality of pairs of matching first descriptors and second descriptors according to the first features and the second features; obtaining position coordinates of the plurality of second descriptors; determining a plurality of corresponding second sub-points in the second point cloud according to the position coordinates of the plurality of second descriptors; deleting the plurality of second sub-points in the second point cloud to form a third point cloud, the third point cloud representing the upper surface of the excess sheet.

[0011] Optionally, the method of determining a plurality of pairs of matching first descriptors and second descriptors according to the first features and the second features includes: constructing a third KD tree according to the first features, the third KD tree indexing the first descriptors in the first features; querying, using the third KD tree, a corresponding nearest neighbor point in the first features for each second descriptor, the nearest neighbor point being a first descriptor in the first features; obtaining the Euclidean distance between the corresponding second descriptor and the nearest neighbor point; determining the corresponding second descriptor and the nearest neighbor point as a plurality of pairs of matching first descriptors and second descriptors if the Euclidean distance is less than a predetermined threshold.

[0012] According to a second aspect of the present application, there is provided an apparatus for identifying an excess sheet, comprising: a point cloud preprocessing module, the point cloud preprocessing module being configured to: scan the surface of a sword grid on a machine tool after the excess sheet has been removed to obtain a first point cloud; extract features from the first point cloud to obtain first features; scan the upper surface of the excess sheet and the surface of the sword grid not covered by the excess sheet to obtain a second point cloud; extract features from the second point cloud to obtain second features; perform feature matching on the first features and the second features to determine the position of the excess sheet on the machine tool.

[0013] According to a third aspect of the present application, there is provided a machine-readable medium having instructions stored thereon, the instructions, when executed by a machine, causing the machine to perform the above-described method of identifying an excess sheet.

[0014] According to a fourth aspect of the present application, there is provided a computing device comprising: a processor; and a memory coupled to the processor configured to store instructions which, when executed by the processor, cause the processor to perform the above-mentioned scrap plate recognition method.

[0015] According to a fifth aspect of the present application, there is provided a computer program product comprising a computer program which, when executed, implements the above-mentioned scrap plate recognition method.

[0016] Compared with the prior art, the technical scheme of the present application has the following beneficial effects: In the scrap plate recognition method provided by the technical scheme of the present application, the first point cloud and the corresponding first features obtained after the scrap plate on the machine tool is emptied are taken as the background, and after the second point cloud and the corresponding second features obtained after the scrap plate is placed on the machine tool are obtained, the first features and the second features are matched to determine the position of the scrap plate on the machine tool. Since the first point cloud is used to represent the surface of the sword grid on the machine tool, and the second point cloud is used to represent the upper surface of the scrap plate and the surface of the sword grid not covered by the scrap plate, the position of the scrap plate can be accurately identified by comparing the first features corresponding to the first point cloud and the second features corresponding to the second point cloud, that is, the scrap plate recognition method can effectively filter out the complex slag background, improve the quality of the segmented scrap plate point cloud, thereby reducing the difficulty of subsequent point cloud segmentation and contour extraction tasks, and ultimately achieving the purpose of improving the algorithm performance and work efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 is a side view structural schematic diagram of an ideal state of a scrap plate and a sword grid; Figure 2 is a point cloud of an ideal state of a scrap plate and a sword grid obtained after scanning the upper surface of the scrap plate and the sword grid not covered by the scrap plate in Figure 1 Figure 3 is a side view structural schematic diagram of a scrap plate and a sword grid with attached slag in an actual processing scene; Figure 4 is a point cloud of a scrap plate and a sword grid with attached slag obtained after scanning the scrap plate and the sword grid with attached slag in Figure 3 Figure 5 is a flowchart of the scrap plate recognition method provided by the embodiment of the present application; Figure 6 ​​is a flowchart of a method for determining the position of a scrap plate on a machine tool based on a first feature and a second feature according to an embodiment of the present application; Figure 7 is a flowchart of a method for determining a plurality of pairs of matched first descriptors and second descriptors according to an embodiment of the present application. DETAILED DESCRIPTION

[0018] As described in the background, it is difficult to accurately identify the scrap plate when the scrap plate is placed on the machine tool with a large amount of slag accumulated. The following will be described in detail in combination with Figures 1 to 4 .

[0019] In one scrap plate identification method, the following steps are included: Step S1, scanning the upper surface of the scrap plate and the sword grid surface not covered by the scrap plate to obtain a point cloud, the point cloud including height information, curvature information and angle information characterizing the upper surface of the scrap plate and the sword grid surface not covered by the scrap plate; Step S2, filtering out the point cloud with height information less than a preset height threshold to obtain a first filtered point cloud; Step S3, according to the curvature information and the angle information of the first filtered point cloud, screening out a second filtered point cloud, the second filtered point cloud being used to represent the upper surface of the scrap plate.

[0020] The above steps distinguish and identify the scrap plate through the height information, angle information and curvature information of the point cloud.

[0021] In an ideal state, please refer to Figure 1 and Figure 2 , Figure 1 is a side view structural schematic diagram of an ideal state of a scrap plate and a sword grid, Figure 2 is a point cloud of the scrap plate and the sword grid in an ideal state obtained by scanning the upper surface of the scrap plate and the sword grid not covered by the scrap plate in Figure 1 . In an ideal state, due to the thickness of the scrap plate itself, the height information of the point cloud representing the upper surface of the scrap plate is greater than the height information of the point cloud representing the surface of the sword grid. Further, since the feature of the scrap plate is high planarity and small curvature, and the feature of the sword grid is large curvature and the plane formed has an angle with the horizontal direction, the point cloud representing the upper surface of the scrap plate and the point cloud representing the surface of the sword grid can be further distinguished through the angle information and the curvature information of the point cloud.

[0022] However, in an actual processing scene, please refer to Figure 3 and Figure 4 , Figure 3 is a side view structural schematic diagram of a scrap plate and a sword grid with slag attached in an actual processing scene, Figure 4 is a point cloud of the scrap plate and the sword grid with slag attached in an actual processing scene obtained by scanning the upper surface of the scrap plate and the sword grid not covered by the scrap plate inFigure 3 The point clouds obtained by scanning the scrap plate and the slag-coated grating are as follows: Because the slag continuously accumulates and settles over time, the height information of the point cloud representing the slag increases. Therefore, it is difficult to distinguish the height information of the point cloud representing the upper surface of the scrap plate from that representing the surface of the grating. Furthermore, due to the continuous accumulation and settling of slag over time, in extreme cases, a relatively flat continuous area may form on the grating, causing a significant reduction in the curvature difference between the point cloud representing the grating surface and the point cloud representing the scrap plate. Thus, it is difficult to distinguish the point cloud representing the upper surface of the scrap plate from that representing the surface of the grating using only the angle and curvature information.

[0023] In view of this, the present invention creatively proposes a method for identifying surplus material plates, comprising: After clearing the residual material plate on the machine tool, scan the surface of the sword grid on the machine tool after clearing the residual material plate to obtain the first point cloud; Feature extraction is performed on the first point cloud to obtain the first feature; After placing the scrap plate on the machine tool, scan the upper surface of the scrap plate and the surface of the sword grid not covered by the scrap plate to obtain the second point cloud. Feature extraction is performed on the second point cloud to obtain the second feature; Feature matching is performed on the first feature and the second feature to determine the position of the scrap plate on the machine tool.

[0024] Since the first point cloud symbolizes the surface of the sword grid on the machine tool, and the second point cloud symbolizes the upper surface of the scrap plate and the surface of the sword grid not covered by the scrap plate, the position of the scrap plate can be accurately identified by comparing the first feature corresponding to the first point cloud and the second feature corresponding to the second point cloud. In other words, this scrap plate identification method can effectively filter out complex slag backgrounds and improve the quality of the segmented scrap plate point cloud, thereby reducing the difficulty of subsequent point cloud segmentation and contour extraction tasks, and ultimately achieving the goal of improving algorithm performance and work efficiency.

[0025] To make the above-mentioned objects, features, and beneficial effects 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 are within the scope of protection of the present invention. The terms "first," "second," "third," etc., in the specification, claims, and accompanying drawings of the present invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the present invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] Please refer to Figure 5 This invention provides a method for identifying surplus material boards, comprising: Step S1: After clearing the residual material plate on the machine tool, scan the surface of the sword grid on the machine tool after clearing the residual material plate to obtain the first point cloud. Step S2: Extract features from the first point cloud to obtain the first feature; Step S3: After placing the scrap plate on the machine tool, scan the upper surface of the scrap plate and the surface of the sword grid not covered by the scrap plate to obtain the second point cloud. Step S4: Extract features from the second point cloud to obtain the second feature; Step S5: Perform feature matching on the first feature and the second feature to determine the position of the scrap plate on the machine tool.

[0027] The following describes the steps in the method for identifying scrap plates.

[0028] In step S1, the first point cloud represents the set of points on the surface of the sword grid on the machine tool, and each point includes corresponding three-dimensional spatial position information.

[0029] In step S2, the method for extracting features from the first point cloud and obtaining the first feature includes: step S21, obtaining the first feature based on the first point cloud using the FPFH Estimation method.

[0030] The first point cloud includes several first sub-points.

[0031] In this embodiment, step S21 specifically includes: Step S211: Obtain the normal of each first sub-point; Step S212: Based on the first point cloud, construct the first KD tree. The first KD tree is used to index the first sub-point in the first point cloud. Step S213: Based on the first KD tree and the normals of each first child point, perform a nearest neighbor search on the first point cloud to construct the first feature.

[0032] Furthermore, before executing step S21, the method includes: step S6, performing voxel filtering on the first point cloud to obtain a first sub-point cloud, wherein the number of first sub-points in the first sub-point cloud is less than the number of first sub-points in the first point cloud.

[0033] Voxel filtering divides the space into uniform voxels, with each voxel retaining one representative point (such as the centroid). This can compress the number of first sub-points while keeping the overall spatial structure of the first point cloud unchanged, thereby reducing the computational load in subsequent processes and saving a lot of computing power.

[0034] Furthermore, after executing step S6, the process also includes: step S7, using the ISS Keypoint 3D method to extract several first sub-points from the first sub-point cloud as several first keypoints, where the number of first keypoints is less than the number of first sub-points in the first sub-point cloud.

[0035] Among them, the ISS Keypoint 3D method analyzes the local neighborhood inertia tensor of the first sub-point and selects the first sub-point with "maximum local curvature and strongest geometric significance" as the first keypoint. This reduces the number of first sub-points with "low geometric information", thereby improving the accuracy of the first feature obtained by the FPFH Estimation method while further saving computing power.

[0036] In step S3, the second point cloud represents the set of points on the upper surface of the scrap plate and the upper surface of the sword grid on the machine tool that is not covered by the scrap plate, and each point includes corresponding three-dimensional spatial position information.

[0037] In step S4, feature extraction is performed on the second point cloud to obtain the second feature. The method includes: step S41, obtaining the second feature based on the second point cloud using the FPFH Estimation method.

[0038] The second point cloud includes several second sub-points.

[0039] In this embodiment, step S41 specifically includes: Step S411: Obtain the normal of each second sub-point; Step S412: Based on the second point cloud, construct a second KD tree. The second KD tree is used to index the second sub-points in the second point cloud. Step S413: Based on the second KD tree and the normals of each second sub-point, perform a nearest neighbor search on the second point cloud to construct the second feature.

[0040] Furthermore, before executing step S41, the process includes: step 8, performing voxel filtering on the second point cloud to obtain a second sub-point cloud, wherein the number of second sub-points in the second sub-point cloud is less than the number of second sub-points in the second point cloud. Its function is the same as step S6, and will not be repeated here.

[0041] Furthermore, after executing step S8, the process also includes step S9, which uses the ISS Keypoint 3D method to extract several second sub-points from the second sub-point cloud as several second keypoints. The number of second keypoints is less than the number of second sub-points in the second sub-point cloud. Its function is the same as step S7, and will not be described again here.

[0042] Please refer to Figure 6 In step S5, the first feature includes several first descriptors, the second feature includes several second descriptors, and the method for feature matching of the first and second features to determine the position of the scrap plate on the machine tool includes: Step S51: Based on the first feature and the second feature, determine several pairs of matching designated first descriptors and designated second descriptors; Step S52: Obtain the position coordinates of several specified second descriptors; Step S53: Based on the position coordinates of several specified second descriptors, determine several corresponding specified second sub-points among several second sub-points; Step S54: Delete several designated second sub-points from the second point cloud to form a third point cloud, which is used to symbolize the upper surface of the remaining material plate.

[0043] Since the matching designated first descriptor and designated second descriptor are the matching parts of the first feature and the second feature, that is, the first descriptor and the second descriptor in the first feature and the second feature symbolize the upper surface of the sword grid of the unobstructed machine tool, after deleting the second sub-point corresponding to the designated second descriptor in the second point cloud, the remaining third point cloud symbolizes the upper surface of the scrap plate.

[0044] Please refer to the following: Figure 7 Step S51 specifically includes: Step S511: Based on the first feature, construct a third KD tree. The third KD tree is used to index the first descriptor in the first feature. Step S512: Using the third KD tree, query the nearest neighbor of each second descriptor in the first feature. The nearest neighbor is a first descriptor in the first feature. Step S513: Obtain the Euclidean distance between the corresponding second descriptor and the nearest neighbor. Step S514: The corresponding second descriptor and nearest neighbor point with a Euclidean distance less than a preset threshold are determined as several pairs of matching designated first descriptors and designated second descriptors.

[0045] In summary, in the scrap plate identification method provided by this invention, a first point cloud and its corresponding first feature are obtained after clearing the scrap plate from the machine tool, serving as the background. A second point cloud and its corresponding second feature are obtained after placing the scrap plate on the machine tool. Feature matching is then performed on the first and second features to determine the position of the scrap plate on the machine tool. Since the first point cloud symbolizes the surface of the slag grid on the machine tool, and the second point cloud symbolizes the upper surface of the scrap plate and the surface of the slag grid not covered by the scrap plate, the position of the scrap plate can be accurately identified by comparing the first feature corresponding to the first point cloud and the second feature corresponding to the second point cloud. In other words, this scrap plate identification method can effectively filter out complex slag backgrounds, improve the quality of the segmented scrap plate point cloud, thereby reducing the difficulty of subsequent point cloud segmentation and contour extraction tasks, ultimately improving algorithm performance and work efficiency.

[0046] This invention also provides a device for identifying scrap boards, including a point cloud preprocessing module.

[0047] The point cloud preprocessing module is used for: Scan the surface of the sword grid on the machine tool after the residual material plate has been cleared to obtain the first point cloud; Feature extraction is performed on the first point cloud to obtain the first feature; Scan the upper surface of the scrap plate and the surface of the sword grid not covered by the scrap plate to obtain a second point cloud; Feature extraction is performed on the second point cloud to obtain the second feature; Feature matching is performed on the first feature and the second feature to determine the position of the scrap plate on the machine tool.

[0048] This invention also provides a machine-readable medium storing instructions that, when executed by a machine, cause the machine to perform the aforementioned waste material board identification method.

[0049] This invention also provides a computing device, including a processor and a memory, the memory being coupled to the processor and configured to store instructions, which, when executed by the processor, cause the processor to perform the aforementioned surplus board identification method.

[0050] This invention also provides a computer program product, including a computer program that, when executed, implements the above-described method for identifying surplus material boards.

[0051] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. A method for identifying surplus material plates, characterized in that, include: After clearing the residual material plate on the machine tool, scan the surface of the sword grid on the machine tool after clearing the residual material plate to obtain the first point cloud; Feature extraction is performed on the first point cloud to obtain the first feature; After placing the scrap plate on the machine tool, scan the upper surface of the scrap plate and the surface of the sword grid not covered by the scrap plate to obtain the second point cloud. Feature extraction is performed on the second point cloud to obtain the second feature; Feature matching is performed on the first feature and the second feature to determine the position of the scrap plate on the machine tool.

2. The method for identifying surplus material plates according to claim 1, characterized in that, The method for extracting features from the first point cloud to obtain the first feature includes: obtaining the first feature based on the first point cloud using the FPFH Estimation method; And / or, the method for extracting features from the second point cloud to obtain the second feature includes: obtaining the second feature based on the second point cloud using the FPFH Estimation method.

3. The method for identifying surplus material plates according to claim 2, characterized in that, The first point cloud includes several first sub-points, and the method for obtaining the first feature by means of the FPFH Estimation method based on the first point cloud includes: obtaining the normal of each first sub-point; constructing a first KD tree based on the first point cloud, the first KD tree being used to index the first sub-points in the first point cloud; and performing a nearest neighbor search on the first point cloud based on the first KD tree and the normal of each first sub-point to construct the first feature. And / or, the second point cloud includes a plurality of second sub-points, and, based on the second point cloud, the method for obtaining the second feature by the FPFHEstimation method includes: obtaining the normal of each second sub-point; constructing a second KD tree based on the second point cloud, the second KD tree being used to index the second sub-points in the second point cloud; and performing a nearest neighbor search on the second point cloud based on the second KD tree and the normal of each second sub-point to construct the second feature.

4. The method for identifying surplus material plates according to claim 2, characterized in that, The first point cloud includes a plurality of first sub-points, and before obtaining the first feature by the FPFH Estimation method based on the first point cloud, the method further includes: performing voxel filtering on the first point cloud to obtain a first sub-point cloud, wherein the number of first sub-points in the first sub-point cloud is less than the number of first sub-points in the first point cloud. And / or, the second point cloud includes a plurality of second sub-points, and, before obtaining the second feature by the FPFHEstimation method based on the second point cloud, the method further includes: performing voxel filtering on the second point cloud to obtain a second sub-point cloud, wherein the number of second sub-points in the second sub-point cloud is less than the number of second sub-points in the second point cloud.

5. The method for identifying surplus material plates according to claim 4, characterized in that, After performing voxel filtering on the first point cloud to obtain the first sub-point cloud, the method further includes: extracting several first sub-points from the first sub-point cloud as several first key points using the ISS Keypoint 3D method, wherein the number of first key points is less than the number of first sub-points in the first sub-point cloud. And / or, after performing voxel filtering on the second point cloud to obtain the second sub-point cloud, the method further includes: extracting several second sub-points from the second sub-point cloud as several second keypoints using the ISS Keypoint3D method, wherein the number of second keypoints is less than the number of second sub-points in the second sub-point cloud.

6. The method for identifying surplus material plates according to any one of claims 1 to 5, characterized in that, The second point cloud includes several second sub-points, the first feature includes several first descriptors, the second feature includes several second descriptors, and the method for feature matching of the first feature and the second feature to determine the position of the scrap plate on the machine tool includes: Based on the first feature and the second feature, determine several pairs of matching designated first descriptors and designated second descriptors; Obtain the position coordinates of several specified second descriptors; Based on the position coordinates of several specified second descriptors, a number of corresponding specified second sub-points are determined among several second sub-points; Several designated second sub-points are deleted from the second point cloud to form a third point cloud, which is used to represent the upper surface of the remaining material plate.

7. The method for identifying surplus material plates according to claim 6, characterized in that, A method for determining several pairs of matching first and second descriptors based on the first feature and the second feature includes: Based on the first feature, a third KD tree is constructed, which is used to index the first descriptor in the first feature; Using the third KD tree, query the nearest neighbor of each second descriptor in the first feature, where the nearest neighbor is one of the first descriptors in the first feature; Obtain the Euclidean distance between the corresponding second descriptor and the nearest neighbor; The corresponding second descriptors and nearest neighbors whose Euclidean distance is less than a preset threshold are determined as several pairs of matching designated first descriptors and designated second descriptors.

8. A device for identifying scrap plates, characterized in that, include: Point cloud preprocessing module, the point cloud preprocessing module is used for: Scan the surface of the sword grid on the machine tool after the remaining material plate has been cleared to obtain the first point cloud; Feature extraction is performed on the first point cloud to obtain the first feature; Scan the upper surface of the scrap plate and the surface of the sword grid not covered by the scrap plate to obtain a second point cloud; Feature extraction is performed on the second point cloud to obtain the second feature; Feature matching is performed on the first feature and the second feature to determine the position of the scrap plate on the machine tool.

9. A machine-readable medium having instructions stored thereon, characterized in that, When executed by a machine, the instructions cause the machine to perform the method as described in any one of claims 1 to 7.

10. A computing device, characterized in that, include: processor; as well as A memory coupled to the processor is configured to store instructions that, when executed by the processor, cause the processor to perform the method as described in any one of claims 1 to 7.

11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed, it implements the method as described in any one of claims 1 to 7.

Images