Complex automobile part edge gap identification and size detection method and system

Abstract

The invention provides a complex automobile part edge gap identification and size detection method and system, and the method comprises the steps: carrying out the clustering segmentation of the preprocessed point cloud sample data of a complex automobile part, and obtaining the three-dimensional point cloud data of the part; extracting the three-dimensional point cloud data of the component by using an expansion convex hull method to obtain edge points; fitting the edge points on the fitting plane, calculating a first distance between the mathematical model and the edge points to obtain defect depth data, and determining potential notch defects on the parts in the complex automobile parts based on the first distance; screening the first distance based on a judgment threshold to obtain a defect position set; edge points in the defect position set are processed based on a screening threshold value, and angular points are obtained; and calculating the distance from each angular point to the fitting plane, determining the width data of the notch defect, and integrating the width data and the depth data to obtain the size data of the complex automobile part. The defect size of the surface of the part can be accurately obtained, and the defect detection efficiency and accuracy of the part are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of parts detection, and in particular to a method and system for edge notch recognition and size detection of complex automobile parts. Background Art

[0002] With the rapid development of the automobile industry, auto parts are not only the cornerstone of the entire vehicle, but also the key factor to ensure driving safety, improve driving experience and extend vehicle service life. Therefore, the quality monitoring of parts, especially the accurate detection of surface defects, has become an indispensable part of the automobile manufacturing process; in modern automobile design, the complexity of parts is increasing day by day, from sophisticated electronic components to complex mechanical structures, each part carries specific functions and safety responsibilities. Although this design progress has brought a leap in performance, it has also brought unprecedented challenges to quality inspection.

[0003] At present, the inspection of automobile parts is often done manually. Although traditional manual inspection methods can complete the inspection of automobile parts, their limitations are becoming increasingly obvious when faced with the current highly complex automobile parts. First, manual inspection is done by inspectors observing the surface of parts and measuring the size of parts with the naked eye with the assistance of measuring and magnifying tools. For complex automobile parts, inspectors need to repeat inspections and measurements many times, which leads to low inspection efficiency and difficulty in meeting the needs of modern automobile production lines for rapid and continuous production. Second, during the manual inspection process, the inspection results are easily affected by the subjective judgment, fatigue level and professional skill level of the inspectors, resulting in inconsistent inspection standards and a high risk of missed inspections and false inspections. Especially when dealing with parts with complex geometric shapes or subtle features, manual inspections often find it difficult to accurately capture all defects, such as tiny cracks, surface flaws or uneven materials. If these defects are not discovered in time, they may cause serious safety hazards in long-term use. Summary of the invention

[0004] In order to solve the problem that the existing manual inspection of complex parts is inefficient and easy to miss, the present invention provides a method and system for edge notch recognition and size detection of complex automobile parts.

[0005] To achieve the above object, the present invention provides the following technical solutions: The present invention proposes a method for edge notch recognition and size detection of complex automobile parts, comprising the following steps: Cluster and segment the pre-processed point cloud sample data of complex automobile parts to obtain the three-dimensional point cloud data of each component in the complex automobile parts; Extracting the three-dimensional point cloud data of the component using an expansion convex hull method to obtain edge points; Fitting all the edge points to a fitting plane with the smallest vertical distance to the edge points by the least square method, establishing a mathematical model of the component surface, and calculating a first distance between the mathematical model and the edge points to obtain defect depth data, and determining potential notch defects on the component in the complex automotive part based on the first distance; The first distance is screened based on a preset judgment threshold to obtain a defect position set; The edge points in the defect position set are screened based on a preset screening threshold, and smoothing detection is performed to obtain corner points; The distance from each corner point to the fitting plane is calculated to obtain a third distance, width data of the notch defect is determined based on the third distance, and the width data and the depth data are integrated to obtain the size data of the complex automobile part.

[0006] Preferably, the clustering and segmenting of the pre-processed point cloud sample data of complex automobile parts comprises: Obtain 3D point cloud data of complex automotive parts; Preprocessing the three-dimensional point cloud data to obtain point cloud sample data; Set a minimum distance, obtain any two points in the point cloud sample data, and calculate the distance between any two points; Comparing the distance between the two points with the minimum distance, if the distance between the two points is less than the minimum distance, assigning the two points to the same cluster; All points in the point cloud sample data are segmented to obtain cluster segmented point cloud data.

[0007] Preferably, the three-dimensional point cloud data is preprocessed to obtain point cloud sample data, including: Divide the three-dimensional point cloud data into a plurality of uniform three-dimensional voxel grids , and calculate each of the three-dimensional voxel grids All points within The mean , integrating all the means Get downsampled point cloud data; Statistical filtering is used to remove noise from the downsampled point cloud data to obtain filtered point cloud data, which is the point cloud sample data.

[0008] Preferably, removing noise from the downsampled point cloud data using statistical filtering comprises: Calculate each point in the downsampled point cloud data The mean distance of the points in the neighborhood of and standard deviation threshold ; Based on the distance mean and the standard deviation threshold Filter the points in the downsampled point cloud data and retain the points within the preset search radius whose average distance to the neighborhood points is less than the standard deviation threshold The remaining points are marked as outliers.

[0009] Preferably, screening the first distance based on a preset judgment threshold includes: Preset judgment threshold; comparing the first distance with the judgment threshold; If the first distance is greater than the judgment threshold, the position of the edge point corresponding to the first distance is marked to obtain the defect position.

[0010] Preferably, screening edge points in the defect position set based on a preset screening threshold comprises: The edge points corresponding to the defect position set are sorted from near to far based on the minimum defect size acceptable to the workpiece as a base point, to obtain an ordered point set of edge points; Preset screening thresholds; Obtain a second distance in the X-axis direction from a corresponding edge point in the edge point ordered point set; Compare the second distance with a screening threshold, extract the corresponding edge points whose second distance is greater than the screening threshold, and obtain four edge points according to the arrangement order of the edge point ordered point set to obtain key points; After edge extraction is performed on the four key points, smoothing detection processing is performed to obtain corner points.

[0011] Preferably, the edge extraction of the key points includes: Perform edge extraction on the four key points, extract points adjacent to the key points, and form a local edge point set; Smoothing detection is performed on all local edge points in the local edge point set to obtain corner points.

[0012] Preferably, the smoothing detection process includes: Performing Gaussian smoothing on all local edge points in the local edge point set to obtain smoothed points; Harris corner point detection is used to extract the smoothed points to obtain corner points.

[0013] Preferably, determining the width data of the notch defect based on the third distance comprises: All the third distances are compared, two corner points corresponding to the smallest value in the third distances are selected, and the distance between the two corner points is measured to obtain width data.

[0014] The present invention proposes a complex automobile part edge notch recognition and size detection system, which is used to implement the above method, including: The preprocessing unit is configured as follows: It is used to cluster and segment the pre-processed point cloud sample data of complex automobile parts to obtain the three-dimensional point cloud data of each component in the complex automobile parts; The first processing unit is configured as follows: Used to fit all edge points to a fitting plane with the smallest vertical distance to the edge points by the least square method, establish a mathematical model of the component surface, and calculate a first distance between the mathematical model and the edge points to obtain defect depth data, and determine potential notch defects on the components in the complex automotive parts based on the first distance; The first screening unit is configured to: Used to screen the first distance based on a preset judgment threshold to obtain a defect position set; The second screening unit is configured to: It is used to filter edge points in the defect position set based on a preset screening threshold, and perform smoothing detection processing to obtain corner points; The second processing unit is configured as follows: Used to calculate the distance from each corner point to the fitting plane to obtain a third distance, determine the width data of the notch defect based on the third distance, and integrate the width data and the depth data to obtain the size data of the complex automobile part; Output unit, configured as: Used to output dimensional data of complex automotive parts.

[0015] Compared with the prior art, the present invention has the following beneficial technical effects: The present invention proposes a method for edge notch recognition and size detection of complex automobile parts. The method utilizes the high precision and comprehensive detection capability of three-dimensional point cloud technology to effectively overcome the limitations of traditional detection methods. The point cloud sample data of complex automobile parts is segmented into three-dimensional point cloud data of each component through clustering segmentation technology, which provides an accurate data basis for subsequent edge extraction and defect identification. The edge points are extracted by the expansion convex hull method, which effectively captures the subtle features of the component surface and provides key information for accurate defect identification. The edge points are fitted to the optimal fitting plane by the least squares method, which not only establishes a mathematical model of the component surface, but also calculates the distance from the edge point to the plane. The first distance accurately identifies potential notch defects. The preset judgment threshold further screens the first distance, ensures the accurate positioning of the defect position, and forms a defect position set. On the basis of the defect position set, the corner points are extracted through the preset screening threshold and smoothing detection processing, and the third distance from the corner point to the fitting plane is calculated, which directly reflects the width data of the notch defect. Combined with the previous depth data, complete size data is formed. The method of the present invention accurately captures and calculates the defect size of the component surface, improves the defect detection efficiency and accuracy of complex automotive parts, and also realizes the accurate measurement of the width and depth of the notch defect, which can improve production efficiency.

[0016] Furthermore, this method calculates the distance between any two points in the point cloud sample data and compares it with the preset minimum distance, intelligently assigns points with a distance less than the threshold to the same cluster, ensuring that points with similar features are classified together, providing a clear and orderly data structure for subsequent edge extraction and defect identification. This segmentation process is performed on all points to ultimately obtain point cloud data after clustering segmentation. These data are not only more structured, but also significantly reduce noise and redundant information, thereby improving the accuracy and efficiency of subsequent analysis. This clustering segmentation method lays a solid foundation for edge notch identification and size detection of complex automotive parts, and helps to achieve more accurate and reliable quality inspection.

[0017] Furthermore, this method achieves downsampling of point cloud data by dividing the three-dimensional voxel grid into uniform grids and calculating the mean of each point in the grid, effectively reducing the amount of data while retaining key features. Then, the statistical filtering method is used to remove noise from the downsampled point cloud data, specifically by calculating the mean and standard deviation of each point and its neighborhood, and screening based on the set first threshold: if the distance between a point and the mean of its neighborhood is greater than a set multiple of the standard deviation (such as twice the standard deviation), it is considered a noise point and removed, thereby obtaining purer filtered point cloud data, namely point cloud sample data, which significantly improves the quality of point cloud data, provides more accurate and reliable input for subsequent edge gap recognition and size detection, and further enhances the accuracy and robustness of the overall method.

[0018] Furthermore, by presetting a judgment threshold and comparing the calculated first distance (i.e., the distance from the edge point to the fitting plane) with the threshold, the method can intelligently identify edge points that are beyond the normal range. These points are regarded as potential defect locations, which improves the accuracy of defect identification and ensures that only points that truly have defect characteristics are marked, avoiding false alarms and missed alarms, and providing a reliable basis for further analysis and processing of subsequent defect locations, thereby improving the efficiency and reliability of the overall detection process.

[0019] Furthermore, this method forms an ordered point set by sorting the edge points in the defect position set by distance, which is conducive to the orderly progress of subsequent analysis. A screening threshold is preset, and screening is performed based on the spacing of the edge points in the X-axis direction, and four key points with a spacing greater than the threshold are extracted, which effectively reduces redundant information and focuses on the edge points that are most likely to constitute a notch defect; edge extraction is performed on these four key points to form a local edge point set, and smoothing detection processing is performed on them, such as Gaussian smoothing and Harris corner point detection, to accurately identify the corner points; corner points are important features that constitute notch defects, and their accurate extraction is crucial for the subsequent measurement of width data; finally, by comparing the third distances of all corner points to the fitting plane, two corner points corresponding to the minimum value are selected, and their distance is measured to obtain the width data of the notch defect, which makes full use of the position information of the corner points and ensures the accuracy of the width data. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A schematic diagram of a flow chart of a method for edge notch recognition and size detection of complex automobile parts proposed by the present invention; Figure 2 This is one of the schematic diagrams of complex automobile parts in an embodiment of the present invention; Figure 3 This is a second schematic diagram of a complex automobile part in an embodiment of the present invention; Figure 4 This is the third schematic diagram of a complex automobile part in an embodiment of the present invention; Figure 5 Schematic diagram of a notch defect of a complex automobile part in an embodiment of the present invention; Figure 6 for Figure 5 The enlarged schematic diagram of the red frame in the middle; Figure 7 A schematic diagram of using a line laser scanner to perform a comprehensive scan of an automobile part in an embodiment of the present invention; Figure 8 Schematic diagram of three-dimensional point cloud data before downsampling in an embodiment of the present invention; Fig. 9 A schematic diagram of downsampling of three-dimensional point cloud data in an embodiment of the present invention; Fig.10 Schematic diagram of down-sampled point cloud data before filtering in an embodiment of the present invention; Fig.11 Schematic diagram of downsampled point cloud data after filtering in an embodiment of the present invention; Fig.12 Schematic diagram of clustering and segmenting point cloud data in an embodiment of the present invention; Fig.13 Point cloud data atlas in the embodiment of the present invention; Fig.14 Schematic diagram of using the expansion convex hull method to obtain edge points in an embodiment of the present invention; Fig.15 It is a schematic diagram of fitting edge points onto a fitting plane in an embodiment of the present invention; Fig.16 is a schematic diagram of a corner point in an embodiment of the present invention; Fig.17 for Fig.16 A magnified schematic diagram of the mid-corner point; Fig.18 is a schematic diagram of a defect position determined by a first distance in an embodiment of the present invention; Fig.19 for Fig.18 Enlarged diagram of the red box in the middle. DETAILED DESCRIPTION

[0021] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0022] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0023] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0024] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, an electrical connection, or a communication; it can be a direct connection, or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0025] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply indicates that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply indicates that the first feature is lower in level than the second feature.

[0026] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0027] See also Figure 1 The present invention proposes a method for edge notch recognition and size detection of complex automobile parts, which comprises the following steps: Cluster and segment the pre-processed point cloud sample data of complex automobile parts to obtain the three-dimensional point cloud data of each component in the complex automobile parts; Specifically, a line laser scanner is used to comprehensively scan automotive parts to obtain high-precision three-dimensional point cloud data of the surface of the parts; Preprocessing of 3D point cloud data includes: Divide 3D point cloud data into multiple uniform 3D voxel grids , and calculate each 3D voxel grid All points within The mean , with the mean As the downsampled point, we get the downsampled point cloud data, and record the points in the downsampled point cloud data as , reduce the amount of data and improve processing efficiency; Among them, point The mean The calculation process is:

[0028] in, is the point cloud within the voxel grid.

[0029] Use statistical filtering to remove noise from the downsampled point cloud data and calculate the The mean distance between the points in its neighborhood and standard deviation threshold .

[0030] Among them, the distance mean The calculation process is:

[0031] in, is the point in the downsampled point cloud data The corresponding field points, is a point in the neighborhood.

[0032] Standard Deviation The calculation process is:

[0033] in, is the point in the downsampled point cloud data The corresponding field points, is a point in the neighborhood, is the mean distance.

[0034] Based on the distance mean and the standard deviation threshold Filter the points in the downsampled point cloud data and retain the points within the preset search radius whose average distance to the neighborhood points is less than the standard deviation threshold The remaining points are marked as outliers. For example, the search radius r is set to calculate the points in the downsampled point cloud data. The average distance of the neighboring points within the search radius (During this calculation, the points in the downsampled point cloud data is the mean of the above corresponding ), the points in the downsampled point cloud data The average distance of the neighboring points within the search radius With standard deviation threshold For comparison, if the points in the downsampled point cloud data The average distance of the neighboring points within the search radius Less than standard deviation threshold , then it is retained; if the points in the downsampled point cloud data The average distance of the neighboring points within the search radius Standard Deviation Threshold , it is marked as an outlier and discarded; the filtering results are sorted to obtain the filtered point cloud data, which is the point cloud sample data.

[0035] The point cloud sample data is clustered and segmented using a clustering method based on a set distance to obtain the label and cluster number of each point, where the label is the point cloud area to which it belongs, and the cluster number is the number of area divisions of the point cloud; for example, the minimum distance is set to , get any two points in the point cloud sample data, and calculate the distance between any two points , if the distance between two points Less than the minimum distance , the two selected points are assigned to the same cluster, and all points in the point cloud sample data are segmented to obtain multiple different cluster numbers and point labels, and the cluster numbers and point labels are recorded, that is, different point cloud areas are obtained, and the corresponding points in each point cloud area are obtained to obtain cluster segmentation point cloud data, that is, the number of point cloud area divisions. Different components in automobile parts are distinguished by cluster segmentation point cloud data, and the three-dimensional point cloud data are classified based on different components to obtain three-dimensional point cloud data of each component in automobile parts.

[0036] Extracting the three-dimensional point cloud data of the component using an expansion convex hull method to obtain edge points; Map the points in the component's 3D point cloud data to obtain the mapping result of the component's 3D point cloud data , where the coordinate data of the points in the component 3D point cloud data is , get the point cloud data atlas, that is, map the points of the component's 3D point cloud data to the XOY plane, and get the mapping result of the component's 3D point cloud data ; Among them, the points in the component 3D point cloud data are plane mapped to obtain the mapping result of the component 3D point cloud data The process is:

[0037] The obtained component 3D point cloud data mapping results Use the expanded convex hull method to obtain the mapping result The edge points of the component are obtained, and then the shape features of the component are obtained. Among them, the expanded convex hull is a method in computational geometry, which is used to determine the minimum outer enclosing shape of a set of points, that is, the expanded convex hull method is used to generate a polyhedron for the component 3D point cloud data. The vertices of the polyhedron are composed of the points in the component 3D point cloud data, and all points are inside or on the surface of this polyhedron. The vertices of the convex hull are edge points, representing the outer contour of the object. After preprocessing the initially acquired 3D point cloud data, the Graham scanning algorithm is used to find the lower left corner point as the starting point, and the remaining points are traversed according to the polar angle sorting. The initial convex hull is generated according to the position relationship between the point and the convex hull edge; then the expansion radius is determined according to the data characteristics and requirements, and the initial convex hull edge endpoint is used as the center of the circle, and a circle is made with the expansion radius, and the external common tangent is connected to achieve expansion to obtain the expanded convex hull; finally, the expanded convex hull is compared with the original point cloud, and the points are classified according to the position, and the boundary points are extracted to form the edge, and then a better edge curve or surface is obtained through post-processing; the expanded convex hull takes into account the distribution of the entire component 3D point cloud data, and can effectively extract all the edge information of the component.

[0038] The expression of the expanded convex hull is:

[0039] in, Represents the expanded convex hull of the component's 3D point cloud data, Represents a point in the 3D point cloud data of a component, located at In space; represents a d-dimensional Euclidean space, It is the dimension of the space where the point in the component's 3D point cloud data is located. Indicates the weight corresponding to each point in the component's three-dimensional point cloud data; represents the number of points used to construct the convex hull; Represents a point in 3D point cloud data.

[0040] Fitting all edge points to a fitting plane with the smallest vertical distance to the edge points by the least square method, establishing a mathematical model of the component surface, and calculating a first distance between the mathematical model and the edge points to obtain defect depth data, and determining potential notch defects on the component in the complex automotive part based on the first distance; Specifically, the extracted edge points are used to determine a plane equation through the least square method, and the plane equation is solved to obtain a fitting plane, the vertical distance between the fitting plane and the edge points is the smallest, all edge points are fitted to the fitting plane, and a mathematical model of the component surface is established; Calculate the first distance between the mathematical model and each edge point, that is, calculate the first distance between the fitting plane and each edge point. The first distance reflects the degree of deviation of each edge point relative to the surface of the object. Based on the degree of deviation, determine the potential notch defect on the component and obtain the defect depth data. That is, compare the degree of deviation with the preset deviation threshold. If the degree of deviation is less than the deviation threshold, it is judged as a normal part. If the degree of deviation is greater than the deviation threshold, it is marked as a defective part point. Calculate the maximum distance between the defective part point and the surface of the object to obtain the defect depth data; where the deviation threshold is the average value of the distance of the most points in the interval where the point distance value is located from the fitting plane.

[0041] The first distance is screened based on a preset judgment threshold to obtain a defect position set; Specifically, a judgment threshold is preset, and the first distances corresponding to all edge points are judged using the judgment threshold, that is, the calculated first distance is compared with the preset judgment threshold. If the first distance is greater than the judgment threshold, the position of the edge point corresponding to the first distance is marked to obtain the defect position, and all defect positions are integrated to obtain a defect position set; The edge points in the defect position set are screened based on a preset screening threshold, and smoothing detection is performed to obtain corner points; Specifically, the corresponding edge points in the defect position set are sorted from near to far according to the distance from the base point with the minimum defect size acceptable to the workpiece as the base point, to obtain an ordered point set of edge points, a screening threshold is preset, and a second distance in the X-axis direction from the corresponding edge points in the ordered point set of edge points is obtained, the second distance is compared with the screening threshold, and the corresponding edge points whose second distance is greater than the screening threshold are extracted, and four edge points are obtained according to the arrangement positions in the ordered point set of edge points to obtain key points; Perform edge extraction on the four key points, extract the points adjacent to the key points, obtain local edge points, and form a local edge point set, which helps to gain a deeper understanding of the geometric features of these key points and provides a basis for further analysis of the features of defects; All local edge points in the local edge point set are subjected to Gaussian smoothing to obtain smoothed points, so as to reduce the slight fluctuations in the data and improve the data quality. This smoothing process can effectively reduce the impact of noise, making subsequent analysis more reliable and accurate.

[0042] Among them, the process of Gaussian smoothing is:

[0043] in, represents the smoothing point, and Represents the position coordinates of the local edge points, is the standard deviation of the Gaussian distribution, controlling the degree of smoothing.

[0044] Harris corner detection is used to extract the smoothed points to obtain corner points; Among them, the process of Harris corner detection is:

[0045] Where R is the response value, which measures whether each smoothing point is a corner point; M is the autocorrelation matrix, which is used to describe the local gradient information of the smoothing point; det(M) is the determinant of the autocorrelation matrix, which indicates the structural strength of the local area; trace(M) is the trace of the autocorrelation matrix, which indicates the overall smoothness of the local area; is a constant used to control the sensitivity of the detection.

[0046] The distance from each corner point to the fitting plane is calculated to obtain a third distance, width data of the notch defect is determined based on the third distance, and the width data and the depth data are integrated to obtain the size data of the complex automobile part.

[0047] Specifically, the distance from each corner point to the fitting plane is calculated to obtain the third distance, all the third distances are compared, the two corner points corresponding to the smallest value in the third distance are selected, and the distance between the two corner points is measured to obtain the fourth distance, that is, the width data. The width data and the depth data are integrated to obtain a set of dimensional data of complex automotive parts. By using this method to perform multiple operations on the same complex automobile part, multiple sets of dimensional data of complex parts are obtained, and the maximum and minimum values ​​are taken to calculate the difference, thereby obtaining the calculated error range.

[0048] The following is a further explanation of the method in conjunction with the embodiments. Figure 2~Figure 4 As shown, defects on complex automotive parts are Figure 5 and Figure 6 As shown; like Figure 7 As shown in the figure, a line laser scanner is used to comprehensively scan automobile parts to obtain high-precision three-dimensional point cloud data on the surface of the parts. Figure 8 As shown, the 3D point cloud data is divided into multiple uniform 3D voxel grids. , and calculate each 3D voxel grid All points within The mean , with the mean As the downsampled point, Fig. 9 and Fig.10 shown.

[0049] Statistical filtering is used to remove noise from the downsampled point cloud data, that is, for each point in the downsampled point cloud data and the mean of the points in its neighborhood and standard deviation .

[0050] The points in the downsampled point cloud data are screened based on the set first threshold, that is, the first threshold is set to , calculate the points in the downsampled point cloud data The average distance of the neighboring points within the search radius , downsample the points in the point cloud data The average distance of the neighboring points within the search radius With the first threshold For comparison, if the points in the downsampled point cloud data The average distance of the neighboring points within the search radius is less than the first threshold, it is retained; if the point in the downsampled point cloud data The average distance of the neighboring points within the search radius Greater than the first threshold , then it is marked as an outlier and discarded; sorting the screening results to obtain the filtered point cloud data, which is the point cloud sample data. Fig.11 As shown; The filtered point cloud data is clustered and segmented using a clustering method based on a set distance to obtain cluster segmented point cloud data. Different components in the automotive parts can be distinguished by cluster segmented point cloud data, such as Fig.12 As shown, the 3D point cloud data is classified based on different components to obtain the 3D point cloud data of each component in the complex automobile parts.

[0051] The points in the component 3D point cloud data are mapped in a plane to obtain the mapping result of the component 3D point cloud data, such as Fig.13 As shown in the figure, the mapping result of the component 3D point cloud data is used to form a polyhedron using the expansion convex hull method, and all points in the component 3D point cloud data are inside or on the surface of the polyhedron. All vertices in the expansion convex hull are extracted to obtain edge points, such as Fig.14 As shown, the appearance characteristics of the component are obtained; Fit all edge points to the fitting plane, such as Fig.15 As shown, a mathematical model of the component surface is established; Calculate the first distance between the mathematical model and each edge point. The first distance reflects the degree of deviation of each edge point relative to the surface of the object. Compare the degree of deviation with the preset deviation threshold. If the degree of deviation is less than the deviation threshold, it is judged as a normal part. If the degree of deviation is greater than the deviation threshold, it is marked as a defective part point. Calculate the maximum distance between the defective part point and the surface of the object to obtain the defect depth data; wherein the deviation threshold is the average value of the distance from the point to the fitting plane with the most points in the interval where the distance value is located, and obtain the defect depth data. The reference value of the degree of deviation is the average value of the distance from the point to the fitting plane with the most points in the interval where the distance value is located.

[0052] The first distances corresponding to all edge points are judged using a judgment threshold, that is, the calculated first distance is compared with the preset judgment threshold. If the first distance is greater than the judgment threshold, the position of the edge point corresponding to the first distance is marked to obtain the defect position, such as Fig.18 and Fig.19 As shown, all defect locations are integrated to obtain a defect location set; The corresponding edge points in the defect position set are sorted from near to far from the base point based on the minimum defect size acceptable to the workpiece, to obtain an ordered set of edge points, a screening threshold is preset, a second distance in the X-axis direction from the corresponding edge points in the ordered set of edge points is obtained, the second distance is compared with the screening threshold, the corresponding edge points whose second distance is greater than the screening threshold are extracted, and four edge points are obtained according to the arrangement positions in the ordered set of edge points to obtain key points; Perform edge extraction on key points, extract points adjacent to key points, obtain local edge points, and form a local edge point set, which helps to gain a deeper understanding of the geometric features of these key points and provides a basis for further analysis of defect features; Gaussian smoothing is performed on all local edge points in the local edge point set to obtain smoothed points, and Harris corner point detection is used to extract the smoothed points to obtain corner points, such as Fig.16 and Fig.17 As shown in the figure, the distance from each corner point to the fitting plane is calculated to obtain the third distance. All the third distances are compared, and the two corner points with the smallest value in the third distance are selected. The distance between the two corner points is measured to obtain the fourth distance, that is, the width data. By integrating the width data and the depth data, a set of dimension data of complex automotive parts is obtained. By using this method to perform 30 operations on the same complex automotive part, the dimension data of multiple sets of complex parts are obtained. The maximum and minimum values ​​are taken to calculate the difference and the error range of the calculation is obtained, as shown in Table 1.

[0053] From Table 1, it can be clearly seen that this method is capable of identifying and sizing edge gaps of complex automotive parts, and the width deviation of the detected dimensions is 0.18, and the depth deviation is 0.25, which are small deviations. Therefore, the defects and dimensional data detected by this method are accurate with small errors.

[0054] The present invention also proposes a complex automobile part edge notch recognition and size detection system for implementing the above method, comprising a preprocessing unit, a first processing unit, a first screening unit, a second screening unit, a second processing unit and an output unit; The preprocessing unit is configured as follows: It is used to cluster and segment the pre-processed point cloud sample data of complex automobile parts to obtain the three-dimensional point cloud data of each component in the complex automobile parts; The first processing unit is configured as follows: Used to fit all edge points to a fitting plane with the smallest vertical distance to the edge points by the least square method, establish a mathematical model of the component surface, and calculate a first distance between the mathematical model and the edge points to obtain defect depth data, and determine potential notch defects on the components in the complex automotive parts based on the first distance; The first screening unit is configured to: Used to screen the first distance based on a preset judgment threshold to obtain a defect position set; The second screening unit is configured to: It is used to filter edge points in the defect position set based on a preset screening threshold, and perform smoothing detection processing to obtain corner points; The second processing unit is configured as follows: Used to calculate the distance from each corner point to the fitting plane to obtain a third distance, determine the width data of the notch defect based on the third distance, and integrate the width data and the depth data to obtain the size data of the complex automobile part; Output unit, configured as: Used to output dimensional data of complex automotive parts.

[0055] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the attached claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved.

[0056] In addition, it should be understood that although this specification is described in accordance with the implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation modes that can be understood by those skilled in the art. The above content is only to illustrate the technical idea of ​​the present invention, and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution according to the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A method for edge notch recognition and size detection of complex automobile parts, characterized in that: The following steps are involved: Cluster and segment the pre-processed point cloud sample data of complex automobile parts to obtain the three-dimensional point cloud data of each component in the complex automobile parts; Extracting the three-dimensional point cloud data of the component using an expansion convex hull method to obtain edge points; Fitting all the edge points to a fitting plane with the smallest vertical distance to the edge points by the least square method, establishing a mathematical model of the component surface, and calculating a first distance between the mathematical model and the edge points to obtain defect depth data, and determining potential notch defects on the component in the complex automotive part based on the first distance; The first distance is screened based on a preset judgment threshold to obtain a defect position set; The edge points in the defect position set are screened based on a preset screening threshold, and smoothing detection is performed to obtain corner points; The distance from each corner point to the fitting plane is calculated to obtain a third distance, width data of the notch defect is determined based on the third distance, and the width data and the depth data are integrated to obtain the size data of the complex automobile part.

2. A complex automobile part edge notch recognition and size detection method according to claim 1, characterized in that: The clustering and segmenting of the pre-processed point cloud sample data of complex automobile parts includes: Obtain 3D point cloud data of complex automotive parts; Preprocessing the three-dimensional point cloud data to obtain point cloud sample data; Set a minimum distance, obtain any two points in the point cloud sample data, and calculate the distance between any two points; Comparing the distance between the two points with the minimum distance, if the distance between the two points is less than the minimum distance, assigning the two points to the same cluster; All points in the point cloud sample data are segmented to obtain cluster segmented point cloud data.

3. A complex automobile part edge notch recognition and size detection method according to claim 2, characterized in that: Preprocessing the three-dimensional point cloud data to obtain point cloud sample data includes: Divide the three-dimensional point cloud data into a plurality of uniform three-dimensional voxel grids , and calculate each of the three-dimensional voxel grids All points within The mean , integrating all the means Get downsampled point cloud data; Statistical filtering is used to remove noise from the downsampled point cloud data to obtain filtered point cloud data, which is the point cloud sample data.

4. A complex automobile part edge notch recognition and size detection method according to claim 3, characterized in that: Using statistical filtering to remove noise from the downsampled point cloud data includes: Calculate each point in the downsampled point cloud data The mean distance of the points in the neighborhood of and standard deviation threshold ; Based on the distance mean and the standard deviation threshold Filter the points in the downsampled point cloud data and keep the points within the preset search radius whose average distance to the neighborhood points is less than the standard deviation threshold The remaining points are marked as outliers.

5. The method for edge notch recognition and size detection of complex automobile parts according to claim 1, characterized in that: Screening the first distance based on a preset judgment threshold includes: Preset judgment threshold; comparing the first distance with the judgment threshold; If the first distance is greater than the judgment threshold, the position of the edge point corresponding to the first distance is marked to obtain the defect position.

6. A complex automobile part edge notch recognition and size detection method according to claim 1, characterized in that: Screening edge points in the defect position set based on a preset screening threshold includes: The edge points corresponding to the defect position set are sorted from near to far based on the minimum defect size acceptable to the workpiece as a base point, to obtain an ordered point set of edge points; Preset screening thresholds; Obtain a second distance in the X-axis direction from a corresponding edge point in the edge point ordered point set; Compare the second distance with a screening threshold, extract the corresponding edge points whose second distance is greater than the screening threshold, and obtain four edge points according to the arrangement order of the edge point ordered point set to obtain key points; After edge extraction is performed on the four key points, smoothing detection processing is performed to obtain corner points.

7. A complex automobile part edge notch recognition and size detection method according to claim 6, characterized in that: The edge extraction of the key points includes: Perform edge extraction on the four key points, extract points adjacent to the key points, and form a local edge point set; Smoothing detection is performed on all local edge points in the local edge point set to obtain corner points.

8. A complex automobile part edge notch recognition and size detection method according to claim 7, characterized in that: The smoothing detection process comprises: Performing Gaussian smoothing on all local edge points in the local edge point set to obtain smoothed points; Harris corner point detection is used to extract the smoothed points to obtain corner points.

9. The method for edge notch recognition and size detection of complex automobile parts according to claim 1, characterized in that: Determining width data of the notch defect based on the third distance includes: All the third distances are compared, two corner points corresponding to the smallest value in the third distances are selected, and the distance between the two corner points is measured to obtain width data.

10. A complex automobile part edge notch recognition and size detection system, used to implement the method described in any one of claims 1 to 9, characterized in that: include: The preprocessing unit is configured as follows: It is used to cluster and segment the pre-processed point cloud sample data of complex automobile parts to obtain the three-dimensional point cloud data of each component in the complex automobile parts; The first processing unit is configured as follows: Used to fit all edge points to a fitting plane with the smallest vertical distance to the edge points by the least square method, establish a mathematical model of the component surface, and calculate a first distance between the mathematical model and the edge points to obtain defect depth data, and determine potential notch defects on the components in the complex automotive parts based on the first distance; The first screening unit is configured to: Used to screen the first distance based on a preset judgment threshold to obtain a defect position set; The second screening unit is configured to: It is used to filter edge points in the defect position set based on a preset screening threshold, and perform smoothing detection processing to obtain corner points; The second processing unit is configured as follows: Used to calculate the distance from each corner point to the fitting plane to obtain a third distance, determine the width data of the notch defect based on the third distance, and integrate the width data and the depth data to obtain the size data of the complex automobile part; Output unit, configured as: Used to output dimensional data of complex automotive parts.

Images