Production quality inspection method, device and system for special-shaped parts

Abstract

The present application relates to the field of image data processing technology, and specifically to a production quality inspection method, device and system for special-shaped parts, the method comprising: obtaining three-dimensional point cloud data and normal vectors of special-shaped parts, extracting key points and their corresponding normal vectors; clustering key points based on similar features between normal vectors and distance features between key points; performing principal component analysis on key points in each cluster cluster to obtain local shape feature values ​​of each cluster cluster; obtaining adjustment indexes of each key point based on the distance distribution between key points in each cluster cluster and the local shape feature values; adjusting the radius in the alpha shape algorithm to obtain the contour of the special-shaped part, performing similarity analysis with the contour of the standard template special-shaped part, and obtaining the quality inspection results of the special-shaped part. The present application aims to optimize the radius in the alpha shape algorithm according to the characteristics of the special-shaped parts to ensure the effectiveness of boundary extraction and improve the accuracy of the quality inspection results.

Description

Technical Field

[0001] The present application relates to the field of image data processing technology, and in particular to a production quality inspection method, device and system for special-shaped parts. Background Art

[0002] In the field of mechanical manufacturing, special-shaped parts are used to produce customized machine parts to meet the special needs of the machine. Due to their unique shape, special processes and technologies such as laser cutting, CNC machining (milling and turning), and 3D printing are often used in the manufacturing process. When the cutting speed and depth are set unreasonably, or the machine tool performance is unstable, it will cause improper cutting force, forming burrs and flash. These problems affect the surface quality, wear resistance and corrosion resistance of special-shaped parts. Therefore, it is necessary to detect defects on the surface of special-shaped parts during the production and processing process, and then eliminate unqualified products.

[0003] When using the Alpha Shape Algorithm to process point cloud data of special-shaped parts, finding the best radius parameter usually requires experimentation and adjustment. Due to the unique shape of special-shaped parts, the choice of radius parameter will directly affect the accuracy and detail of boundary extraction. If the radius parameter is not set appropriately, it will cause errors in the boundary extraction results, thus affecting the accuracy of subsequent quality inspection results. Summary of the invention

[0004] In view of the above, it is necessary to provide a production quality inspection method, device and system for special-shaped parts to solve the above problems.

[0005] According to one aspect of the present application, a production quality inspection method for special-shaped parts is provided, the method comprising:

[0006] Obtain 3D point cloud data of special-shaped parts;

[0007] The feature point detection algorithm is used for the 3D point cloud data of special-shaped parts to extract key points and their corresponding normal vectors. For special-shaped parts, the key points are clustered based on the similarity features between the normal vectors of the key points and the distance features between the key points.

[0008] Perform principal component analysis on the key points in each cluster to obtain the Gaussian curvature distribution of the key points, and combine the distance between the key points to obtain the local shape feature value of each cluster;

[0009] In each cluster, neighboring key points of each key point are preset, and the adjustment index of each key point is obtained according to the distance distribution between each key point and its neighboring key points in each cluster and the local shape feature value;

[0010] According to the adjustment index of all key points, the radius in the Alpha shape algorithm is adjusted to obtain the contour of the special-shaped part. The similarity analysis is performed with the contour of the standard template special-shaped part to obtain the quality inspection result of the special-shaped part.

[0011] The clustering of key points is specifically as follows:

[0012] Calculate the cosine similarity between the normal vectors of the key points; calculate the Euclidean distance between the key points; use the sum of the normalized value of the cosine similarity and the normalized value of the Euclidean distance as a metric of the clustering algorithm, and use the clustering algorithm in combination with the metric to obtain each cluster cluster.

[0013] The local shape feature values ​​of each cluster are obtained as follows:

[0014] All key points in each cluster are numbered based on Gaussian curvature, and the prominent points in each cluster are screened according to the distance between adjacent numbered key points to obtain the regional smoothness of each cluster;

[0015] The regional smoothness of each cluster is normalized; the ratio of the number of key points of non-salient points in each cluster to the number of all key points is calculated, and the normalized regional smoothness is merged with the ratio to obtain the local shape feature value of each cluster.

[0016] The screening of the outstanding points in each cluster is specifically as follows:

[0017] The key points in each cluster are numbered from small to large according to the Gaussian curvature, and the absolute value of the Gaussian curvature difference between each key point and the adjacent key point is calculated; the key point whose absolute value of the Gaussian curvature difference is greater than the difference threshold is marked as a prominent point.

[0018] The regional smoothness of each cluster is obtained as follows:

[0019] The average level of distances between all numbered adjacent key points in each cluster is obtained, recorded as the first distance; the average level of Gaussian curvature of all key points in each cluster is extracted, recorded as the average curvature; the first distance is merged with the average curvature to obtain the regional smoothness of each cluster.

[0020] Wherein, obtaining the adjustment index of each key point includes:

[0021] In each cluster, the average distance between all key points and their preset number of neighboring key points is obtained, which is recorded as the second distance;

[0022] With each key point as the center of the circle and the preset multiple of the second distance as the neighborhood radius, the number of all key points in the neighborhood of each key point is taken as the local density of each key point;

[0023] According to the second distance of each cluster and the local shape eigenvalue, combined with the local density of each key point, an adjustment index of each key point is obtained; wherein the adjustment index is positively correlated with the second distance and the local density, and negatively correlated with the local shape eigenvalue.

[0024] The adjustment of the radius in the alpha shape algorithm is specifically as follows: Among them, α represents the radius of the key point corresponding to the alpha shape algorithm; Lca represents the adjustment index of the key point; λ represents the preset value; μ represents the average distance between the key point and all its neighboring key points; e represents a natural constant.

[0025] The quality inspection results of the special-shaped parts are obtained as follows:

[0026] All point cloud data are used as input of the Alpha shape algorithm, combined with the radius corresponding to all key points, to obtain the contour of the special-shaped part, and the contour feature extraction algorithm is used to obtain the feature vector of the contour of the special-shaped part; the normalized value of the distance metric between the feature vector of the contour of the special-shaped part and the contour of the standard template special-shaped part is calculated;

[0027] When the normalized value of the distance metric is less than or equal to the preset distance threshold, the contour of the special-shaped part meets the quality inspection standard; otherwise, it does not meet the quality inspection standard.

[0028] According to another aspect of the present application, a production quality inspection system for special-shaped parts is provided, comprising:

[0029] Special-shaped parts point cloud data acquisition module, used to obtain three-dimensional point cloud data of special-shaped parts;

[0030] The special-shaped parts point cloud data processing module is used to use a feature point detection algorithm to extract key points and their corresponding normal vectors from the three-dimensional point cloud data of the special-shaped parts; for special-shaped parts, the key points are clustered based on the similarity characteristics between the normal vectors of the key points and the distance characteristics between the key points; the key points in each cluster are subjected to principal component analysis to obtain the Gaussian curvature distribution of the key points, and the local shape feature values ​​of each cluster are obtained in combination with the distances between the key points; in each cluster, the neighboring key points of each key point are preset, and the adjustment index of each key point is obtained according to the distance distribution between each key point and its neighboring key points in each cluster, combined with the local shape feature values;

[0031] The special-shaped parts contour extraction quality inspection module is used to adjust the radius in the Alpha shape algorithm according to the adjustment index of all key points to obtain the contour of the special-shaped parts, and perform similarity analysis with the contour of the standard template special-shaped parts to obtain the quality inspection results of the special-shaped parts.

[0032] According to another aspect of the present application, a production quality inspection device for special-shaped parts is provided, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements operations corresponding to the above-mentioned production quality inspection method for special-shaped parts when executing the computer program.

[0033] In the above scheme, obtaining the three-dimensional point cloud data of the special-shaped parts is helpful to understand the geometric characteristics of the surface of the special-shaped parts and deal with surface defects; extracting key points and their corresponding normal vectors is helpful to obtain the shape characteristics of the special-shaped parts and improve the analysis efficiency; clustering the key points based on the distance, normal vector and other characteristics of the key points is helpful to more comprehensively understand the structure and characteristics of the special-shaped parts, and cluster the key points with similar structures into a cluster; principal component analysis is performed on the key points in each cluster to obtain the Gaussian curvature distribution of the key points, and the local shape feature values ​​of each cluster are obtained in combination with the distance between the key points. The beneficial effect is that the local features of the special-shaped parts are characterized, and then the adjustment index of each key point in the cluster is constructed, which is convenient for adjusting the radius in the alpha shape algorithm according to the local features of the special-shaped parts, and selecting the appropriate radius according to the point cloud distribution characteristics of the special-shaped parts to obtain a more accurate and effective special-shaped part contour; finally, the contour of the special-shaped part is compared with the contour of the standard template special-shaped part, so that the quality inspection result of the special-shaped part is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flowchart of the steps of the production quality inspection method for special-shaped parts provided in this application;

[0035] Figure 2 A schematic diagram of point cloud data contour extraction before radius adjustment of the Alpha shape algorithm provided in this application;

[0036] Figure 3 A schematic diagram of contour extraction of point cloud data after radius adjustment of the Alpha shape algorithm provided in this application;

[0037] Figure 4 Block diagram of the production quality inspection system for special-shaped parts provided for this application. DETAILED DESCRIPTION

[0038] In the description of the embodiments of the present application, words such as "exemplary", "or", "for example" and the like are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "or", "for example" and the like is intended to present related concepts in a concrete manner.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art in the present application. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application.

[0040] It should also be noted that the terms "first" and "second" in this application and the accompanying drawings are used to distinguish similar objects, rather than to describe a specific order or sequence. The method disclosed in the embodiments of the present application or the method shown in the flowchart includes one or more steps for implementing the method. Without departing from the scope of the present application, the execution order of multiple steps can be interchanged with each other, and some steps can also be deleted.

[0041] See also Figure 1 , which shows a flowchart of a production quality inspection method for special-shaped parts provided by an embodiment of the present application, the method comprising the following steps:

[0042] Step 1: Obtain 3D point cloud data of special-shaped parts.

[0043] Special-shaped parts usually have complex geometric shapes, asymmetry, and many holes. Therefore, in order to accurately obtain the data of these features, a 3D scanner is used to scan the special-shaped parts and obtain the three-dimensional point cloud data of the special-shaped parts.

[0044] The acquired three-dimensional point cloud data is denoised by median filtering, wherein the median filtering algorithm is an existing well-known technology and will not be described in detail in this application. Furthermore, the implementer can select a suitable denoising algorithm to process the point cloud data.

[0045] It should be noted that the 3D point cloud data of special-shaped parts mentioned later are all denoised 3D point cloud data.

[0046] Step 2: Use feature point detection algorithm to extract key points and their corresponding normal vectors from the three-dimensional point cloud data of special-shaped parts; for special-shaped parts, cluster the key points based on the similarity features between the normal vectors of the key points and the distance features between the key points.

[0047] After cutting, stamping or grinding metal processing operations, special-shaped parts often form sharp and irregular protrusions at the edges or corners due to the plastic deformation of the material. These are the burrs and flash defects on the surface of special-shaped parts. Burrs appear as small, sharp thorns with varying sizes and shapes. They are often distributed on the edges of parts and holes, especially where the tool contacts the material during cutting or stamping. Flash is a thin sheet of material on the edge of the part, mainly distributed at the edge and interface of the part. They are sharp and irregular protrusions generated by metal during plastic deformation.

[0048] Based on this, a feature point detection algorithm is used for the three-dimensional point cloud data of special-shaped parts to extract key points and their corresponding normal vectors: This embodiment uses the SIFT3D algorithm to extract key points and their corresponding normal vectors from the three-dimensional point cloud data of special-shaped parts, wherein the SIFT3D algorithm is an existing public technology, and the monitoring threshold of key points in the algorithm is set to 0.1 in this embodiment. The implementer can adjust it according to the actual situation, and this application will not elaborate on it.

[0049] Burrs and flashes form sharp edges and angles on special-shaped parts, which causes significant deviations in the point cloud data normal vectors of these areas, while the surface normal vectors of normal special-shaped parts change smoothly. In addition, the shapes of burrs and flashes are usually irregular, making the point cloud distribution of these defect areas appear more chaotic.

[0050] For special-shaped parts, the key points are clustered based on the similarity features between the normal vectors of the key points and the distance features between the key points: in this embodiment, the cosine similarity between the normal vectors of the key points is calculated; the Euclidean distance between the key points is calculated; the sum of the normalized value of the cosine similarity and the normalized value of the Euclidean distance is used as the metric of the DBSCAN clustering algorithm, and the clustering algorithm is used in combination with the metric to obtain each cluster; wherein the neighborhood radius of the DBSCAN clustering algorithm is the lower quartile of the Euclidean distance between all two key points; the minimum number of neighborhood points is set to 50.

[0051] Step 3: Perform principal component analysis on the key points in each cluster to obtain the Gaussian curvature distribution of the key points, and combine the distance between the key points to obtain the local shape feature value of each cluster.

[0052] Using principal component analysis, we can obtain the Gaussian curvature of each key point in each cluster:

[0053] In this embodiment, the set of all key points in each cluster is used as the input of the principal component analysis to obtain the principal components of each cluster and their corresponding eigenvalues; the Gaussian curvature of each key point is calculated based on the eigenvalues ​​combined with the principal curvature of each key point in the principal component analysis.

[0054] It should be noted that principal component analysis is a well-known technology, and this application will not elaborate on it. The curvature of the key point reflects the local geometric characteristics of each key point, rather than the overall curvature of the entire cluster.

[0055] All key points in each cluster are numbered based on Gaussian curvature, and the prominent points in each cluster are screened according to the distance between adjacent numbered key points to obtain the regional smoothness of each cluster:

[0056] The key points in each cluster are numbered in ascending order according to the Gaussian curvature, and the absolute value of the Gaussian curvature difference between each key point and the next key point is calculated; the key points whose absolute value of the Gaussian curvature difference is greater than the difference threshold are marked as prominent points; the number of prominent points in each cluster is counted; if there are key points with the same Gaussian curvature, they are numbered in ascending order according to the distance from the last key point numbered; the difference threshold in this embodiment is obtained by the Otsu threshold, wherein the Otsu threshold algorithm is a prior art known in the art and will not be elaborated in this application.

[0057] Furthermore, the average level of distances between all numbered adjacent key points in each cluster is obtained, recorded as the first distance; the average level of Gaussian curvature of all key points in each cluster is extracted, recorded as the average curvature; the first distance is merged with the average curvature to obtain the regional smoothness of each cluster.

[0058] In this embodiment, the average levels of multiple variables are reflected by the average value, and the average value of the Euclidean distances between all numbered adjacent key points in each cluster is calculated; the multiple variables are fused by multiplication to obtain the product of the first distance and the average curvature; the negative correlation mapping of the variables is reflected by the reciprocal of the variables, that is, the reciprocal of the product of each cluster is used as the regional smoothness of each cluster.

[0059] It should be understood that the smaller the distance between key points with adjacent numbers in each cluster, the smaller the first distance; if the curvature value of all key points in a cluster is small, this usually means that the area is relatively flat on a local scale, that is, the geometric shape corresponding to the cluster is relatively smooth.

[0060] The regional smoothness of each cluster is normalized; the ratio of the number of key points of non-salient points in each cluster to the number of all key points is calculated, and the normalized regional smoothness is merged with the ratio to obtain the local shape feature value of each cluster.

[0061] In this embodiment, the fusion between variables is in the form of multiplication followed by square root taking, that is, the local shape feature value of each cluster is specifically the product of the ratio and the normalized regional smoothness, and then the square root value is taken.

[0062] It should be understood that when the number of non-protruding key points in each cluster accounts for a large proportion, it means that the area of ​​the special-shaped parts corresponding to the key points in each cluster is flatter; when the regional smoothness of each cluster is greater, it means that the geometric shape corresponding to the cluster has obvious smooth characteristics; and thus the local shape feature value of the special-shaped parts is greater.

[0063] Step 4: In each cluster, preset the neighboring key points of each key point, and obtain the adjustment index of each key point according to the distance distribution between each key point and its neighboring key points in each cluster and the local shape feature value.

[0064] When using the Alpha-shapes algorithm to extract the boundary of the point cloud data of special-shaped parts, the parameter in the algorithm is the radius of the rolling circle, denoted as α. When the α value is small, it can touch more gaps between points and extract more boundary details, but it may cause overfitting and create the risk of judging each point as a boundary point; when the α value is large, the extracted boundary tends to be simple and overall, and some small concave parts or inner boundaries may be ignored. Based on this, in each cluster, the neighboring key points of each key point are preset, and the adjustment index of each key point is obtained according to the distance distribution between each key point and its neighboring key points in each cluster, combined with the local shape feature value:

[0065] In each cluster, the average level of the distance between all key points and a preset number of neighboring key points is obtained, which is recorded as the second distance; with each key point as the center of the circle, a preset multiple of the second distance is used as the neighborhood radius, and the number of all key points in the neighborhood of each key point is used as the local density of each key point; according to the second distance of each cluster and the local shape eigenvalue, combined with the local density of each key point, an adjustment index of each key point is obtained; wherein the adjustment index is positively correlated with the second distance and the local density, and negatively correlated with the local shape eigenvalue.

[0066] In this embodiment, the preset number is 6, and the implementer can adjust it according to actual conditions, and this application does not impose any restrictions on this; the second distance is specifically the average of the distances between all key points and their preset number of neighboring key points; the preset multiple is 3; the adjustment index of each key point is specifically: calculate the product of the local density of each key point and the second distance of the cluster to which it is located; calculate 1 and the sum of the local shape feature values ​​of the cluster to which each key point is located; and use the ratio of the product to the sum as the adjustment index of each key point.

[0067] It should be understood that if the key points in the cluster are sparsely distributed, the larger the neighborhood radius of each key point, the larger the neighborhood radius usually needs to be to ensure that there are enough points in the neighborhood of each key point for effective analysis or calculation; since the key points are sparsely distributed, the smaller the local shape eigenvalue is, the larger the value of the adjustment index is.

[0068] Step 5: According to the adjustment index of all key points, adjust the radius in the alpha shape algorithm to obtain the outline of the special-shaped part.

[0069] The larger the adjustment index is, the larger the α value should be to avoid breaks in the extracted boundaries due to the sparsity of the point cloud. The smaller the adjustment index is, the smaller the α value should be. The α value in the algorithm can be adjusted through the adjustment index of each key point to extract a more accurate and continuous contour, thereby improving the quality of point cloud data processing.

[0070] In this embodiment, the specific formula of the radius of each key point corresponding to the alpha shape algorithm is: Among them, α represents the radius of the key point corresponding to the alpha shape algorithm; Lca represents the adjustment index of the key point; λ represents the preset value, which is 1 and can be set by the implementer according to the actual situation; μ represents the average distance between the key point and all its neighboring key points; e represents a natural constant.

[0071] All point cloud data are used as input of the Alpha shape algorithm, combined with the radius corresponding to all key points, to obtain the contour of the special-shaped part, and the contour feature extraction algorithm is used to obtain the feature vector of the contour of the special-shaped part; the normalized value of the distance measurement between the feature vector of the contour of the special-shaped part and the standard template special-shaped part is calculated. The contour feature extraction algorithm used in this embodiment is the Zernike moment algorithm; the distance measurement between the feature vectors uses the DTW distance; the sigmoid function is used for normalization. When the normalized value of the DTW distance is less than or equal to the preset distance threshold, the contour of the special-shaped part meets the quality inspection standard, otherwise, it does not meet the quality inspection standard. Among them, the distance threshold is set to 0.2; the implementer can adjust it by himself.

[0072] Among them, the schematic diagram of point cloud data contour extraction before radius adjustment of the alpha shape algorithm is as follows Figure 2 As shown; Schematic diagram of point cloud data contour extraction after radius adjustment of the Alpha shape algorithm, as shown Figure 3 shown.

[0073] It should be noted that the Alpha shape algorithm, the Zernike moment algorithm, and the DTW distance are all existing well-known technologies, and this application will not elaborate on them.

[0074] Based on the same concept as the method embodiment of the present application, a production quality inspection system for special-shaped parts is proposed, including:

[0075] Special-shaped parts point cloud data acquisition module, used to obtain three-dimensional point cloud data of special-shaped parts;

[0076] The special-shaped parts point cloud data processing module is used to use a feature point detection algorithm to extract key points and their corresponding normal vectors from the three-dimensional point cloud data of the special-shaped parts; for special-shaped parts, the key points are clustered based on the similarity characteristics between the normal vectors of the key points and the distance characteristics between the key points; the key points in each cluster are subjected to principal component analysis to obtain the Gaussian curvature distribution of the key points, and the local shape feature values ​​of each cluster are obtained in combination with the distances between the key points; in each cluster, the neighboring key points of each key point are preset, and the adjustment index of each key point is obtained according to the distance distribution between each key point and its neighboring key points in each cluster, combined with the local shape feature values;

[0077] The special-shaped parts contour extraction quality inspection module is used to adjust the radius in the Alpha shape algorithm according to the adjustment index of all key points to obtain the contour of the special-shaped parts, and perform similarity analysis with the contour of the standard template special-shaped parts to obtain the quality inspection results of the special-shaped parts.

[0078] Among them, the production quality inspection system block diagram of special-shaped parts is as follows: Figure 4 shown.

[0079] Based on the same concept as the method embodiment of the present application, a production quality inspection device for special-shaped parts is proposed, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the operations corresponding to the above-mentioned production quality inspection method for special-shaped parts are implemented.

[0080] In summary, obtaining three-dimensional point cloud data of special-shaped parts is helpful to understand the geometric characteristics of the surface of special-shaped parts and deal with surface defects; extracting key points and their corresponding normal vectors is helpful to obtain the shape characteristics of special-shaped parts and improve analysis efficiency; clustering key points based on key point distance, normal vector and other features is helpful to more comprehensively understand the structure and characteristics of special-shaped parts, and cluster key points with similar structures into a cluster; principal component analysis is performed on the key points in each cluster to obtain the Gaussian curvature distribution of the key points, and the local shape feature values ​​of each cluster are obtained in combination with the distance between the key points. The beneficial effect is that the local features of the special-shaped parts are characterized, and then the adjustment index of each key point in the cluster is constructed, which is convenient for adjusting the radius in the alpha shape algorithm according to the local features of the special-shaped parts, and selecting the appropriate radius according to the point cloud distribution characteristics of the special-shaped parts to obtain a more accurate and effective special-shaped part contour; finally, the contour of the special-shaped part is compared with the contour of the standard template special-shaped part, so that the quality inspection result of the special-shaped part is more accurate.

[0081] It should be noted that the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the system, method and computer program product according to the embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. In the description corresponding to the flowchart and block diagram in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in an order different from that disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.

[0082] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A production quality inspection method for special-shaped parts, characterized in that: The method comprises the following steps: Obtain 3D point cloud data of special-shaped parts; The feature point detection algorithm is used for the 3D point cloud data of special-shaped parts to extract key points and their corresponding normal vectors. For special-shaped parts, the key points are clustered based on the similarity features between the normal vectors of the key points and the distance features between the key points. Perform principal component analysis on the key points in each cluster to obtain the Gaussian curvature distribution of the key points, and combine the distance between the key points to obtain the local shape feature value of each cluster; In each cluster, neighboring key points of each key point are preset, and the adjustment index of each key point is obtained according to the distance distribution between each key point and its neighboring key points in each cluster and the local shape feature value; According to the adjustment index of all key points, the radius in the Alpha shape algorithm is adjusted to obtain the contour of the special-shaped part. The similarity analysis is performed with the contour of the standard template special-shaped part to obtain the quality inspection result of the special-shaped part.

2. The production quality inspection method for special-shaped parts according to claim 1, characterized in that: The clustering of key points is specifically as follows: Calculate the cosine similarity between the normal vectors of the key points; calculate the Euclidean distance between the key points; use the sum of the normalized value of the cosine similarity and the normalized value of the Euclidean distance as a metric of the clustering algorithm, and use the clustering algorithm in combination with the metric to obtain each cluster cluster.

3. The production quality inspection method of special-shaped parts according to claim 1, characterized in that: The local shape feature values ​​of each cluster are obtained as follows: All key points in each cluster are numbered based on Gaussian curvature, and the prominent points in each cluster are screened according to the distance between adjacent numbered key points to obtain the regional smoothness of each cluster; Normalize the regional smoothness of each cluster; The ratio of the number of key points of non-salient points in each cluster to the number of all key points is calculated, and the normalized regional smoothness is fused with the ratio to obtain the local shape feature value of each cluster.

4. The production quality inspection method of special-shaped parts according to claim 3, characterized in that: The screening of the outstanding points in each cluster is specifically as follows: The key points in each cluster are numbered from small to large according to the Gaussian curvature, and the absolute value of the Gaussian curvature difference between each key point and the adjacent key point is calculated; the key point whose absolute value of the Gaussian curvature difference is greater than the difference threshold is marked as a prominent point.

5. The production quality inspection method of special-shaped parts according to claim 3, characterized in that: The regional smoothness of each cluster is obtained as follows: The average level of distances between all numbered adjacent key points in each cluster is obtained, recorded as the first distance; the average level of Gaussian curvature of all key points in each cluster is extracted, recorded as the average curvature; the first distance is merged with the average curvature to obtain the regional smoothness of each cluster.

6. The production quality inspection method for special-shaped parts according to claim 1, characterized in that: The step of obtaining the adjustment index of each key point includes: In each cluster, the average distance between all key points and their preset number of neighboring key points is obtained, which is recorded as the second distance; With each key point as the center of the circle and the preset multiple of the second distance as the neighborhood radius, the number of all key points in the neighborhood of each key point is taken as the local density of each key point; According to the second distance of each cluster and the local shape eigenvalue, combined with the local density of each key point, an adjustment index of each key point is obtained; wherein the adjustment index is positively correlated with the second distance and the local density, and negatively correlated with the local shape eigenvalue.

7. The production quality inspection method of special-shaped parts according to claim 1, characterized in that: The adjustment of the radius in the alpha shape algorithm is specifically as follows: Among them, α represents the radius of the key point corresponding to the alpha shape algorithm; Lca represents the adjustment index of the key point; λ represents the preset value; μ represents the average distance between the key point and all its neighboring key points; e represents a natural constant.

8. The production quality inspection method for special-shaped parts according to claim 1, characterized in that: The quality inspection results of the special-shaped parts are specifically as follows: All point cloud data are used as input of the Alpha shape algorithm, combined with the radius corresponding to all key points, to obtain the contour of the special-shaped part, and the contour feature extraction algorithm is used to obtain the feature vector of the contour of the special-shaped part; the normalized value of the distance metric between the feature vector of the contour of the special-shaped part and the contour of the standard template special-shaped part is calculated; When the normalized value of the distance metric is less than or equal to a preset distance threshold, the contour of the special-shaped part meets the quality inspection standard; otherwise, it does not meet the quality inspection standard.

9. The production quality inspection system for special-shaped parts is characterized by: The system includes: Special-shaped parts point cloud data acquisition module, used to obtain three-dimensional point cloud data of special-shaped parts; The special-shaped parts point cloud data processing module is used to use a feature point detection algorithm to extract key points and their corresponding normal vectors from the three-dimensional point cloud data of the special-shaped parts; for special-shaped parts, the key points are clustered based on the similarity characteristics between the normal vectors of the key points and the distance characteristics between the key points; the key points in each cluster are subjected to principal component analysis to obtain the Gaussian curvature distribution of the key points, and the local shape feature values ​​of each cluster are obtained in combination with the distances between the key points; in each cluster, the neighboring key points of each key point are preset, and the adjustment index of each key point is obtained according to the distance distribution between each key point and its neighboring key points in each cluster, combined with the local shape feature values; The special-shaped parts contour extraction quality inspection module is used to adjust the radius in the Alpha shape algorithm according to the adjustment index of all key points to obtain the contour of the special-shaped parts, and perform similarity analysis with the contour of the standard template special-shaped parts to obtain the quality inspection results of the special-shaped parts.

10. A production quality inspection device for special-shaped parts, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, operations corresponding to the production quality inspection method for special-shaped parts according to any one of claims 1 to 8 are implemented.

Images