A quality inspection method for automobile taillight mask molding

Abstract

The present invention relates to the field of image processing, and more particularly to a method for quality inspection of automobile taillight mask molding, comprising: collecting and preprocessing three-dimensional point cloud data of the automobile taillight mask; calculating the tail mask structure support of each three-dimensional point cloud; calculating the local curvature of each three-dimensional point cloud; calculating the tail mask structure significance index of each three-dimensional point cloud based on the tail mask structure support and the local curvature; screening the three-dimensional point cloud data of the automobile taillight mask based on the calculated tail mask structure significance index; and matching the screened three-dimensional point cloud data to achieve molding quality inspection of the automobile taillight mask. The present invention screens the three-dimensional point cloud data based on the tail mask structure significance index, retains key feature points, and can quickly and accurately detect the molding quality of the taillight mask.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and more particularly to a quality inspection method for automobile taillight mask molding. Background Art

[0002] The automotive taillight mask is a key functional component in the taillight assembly. As a transparent or translucent plastic protective cover over the taillight bulb, reflector, and circuitry, it simultaneously performs multiple important functions. The taillight mask not only provides physical protection for the internal optical components, preventing environmental corrosion, but also achieves light diffusion and directional reflection through precisely designed optical surfaces, ensuring that the taillight's optical performance meets regulatory requirements. Given the critical role of the taillight mask in safety performance and visual quality, precise testing of its molding quality is of great engineering significance.

[0003] High-precision 3D inspection technology is currently widely used in the industry for quality control. The ICP (Iterative Closest Point) registration algorithm, based on point cloud data, is widely used due to its exceptional measurement accuracy. This algorithm iteratively matches the 3D scan data of the test sample with the standard CAD model to detect shape deviations, providing a reliable quantitative assessment of the molding quality of the taillight fascia.

[0004] In the three-dimensional point cloud data of a car's taillight visor, the point cloud density varies significantly across different regions. It's worth noting that while the number of points in sparse regions is relatively small, due to their spatial distribution, these discrete points have a disproportionately important impact on the representation of the overall structural features. However, traditional point cloud processing methods, which typically employ uniform sampling or global optimization strategies, fail to fully account for this non-uniform density distribution and its impact on structural reconstruction, resulting in distortion or loss of structural features in sparse regions during reconstruction. Summary of the Invention

[0005] In order to solve the above-mentioned technical problem of insufficient recognition accuracy of key areas of the taillight mask, the present invention provides the following technical solution.

[0006] A quality inspection method for automobile taillight mask molding, comprising:

[0007] Collect 3D point cloud data of car taillight mask and perform preprocessing;

[0008] Calculate the tail cover structure significance index, filter the 3D point cloud data of the car tail light cover based on the calculated tail cover structure significance index, and perform ICP matching between the filtered 3D point cloud data and the 3D point cloud data of the designed shape to evaluate the deviation between the actual shape of the car tail light cover and the designed shape, thereby realizing the molding quality inspection of the car tail light cover;

[0009] The calculation process of the tail cover structure significance index is:

[0010] Calculate the tail hood structure support of each three-dimensional point cloud, and the tail hood structure support is used to reflect the degree of change of the tail hood structure after deleting the three-dimensional point cloud; calculate the local curvature of each three-dimensional point cloud; select any three-dimensional point cloud as the target point cloud, and determine the neighboring point set of the target point cloud; calculate the product of the tail hood structure support and the local curvature of each three-dimensional point cloud in the neighboring point set, and find the maximum and minimum values ​​of the product, and use the difference between the product corresponding to the three-dimensional point cloud in the neighboring point set and the minimum value of the product as the first parameter, and use the difference between the maximum value of the product in the neighboring point set and the product corresponding to the three-dimensional point cloud in the neighboring point set as the second parameter, and then sum the ratios of the first parameter to the second parameter of each three-dimensional point cloud in the neighboring point set to obtain the tail hood structure significance index of the target point cloud.

[0011] The present invention first enhances 3D point cloud features by calculating the structural support and local curvature of the 3D point cloud. The structural support reflects the contribution of the point cloud to the overall taillight visor shape, while the local curvature characterizes surface geometric features (such as edges and corners). The product of the structural support and local curvature highlights points with both high geometric characteristic and structural importance. The product of the structural support and local curvature is then combined to calculate the taillight visor structural significance index of the 3D point cloud, quantifying the structural characteristics of the taillight visor. The 3D point cloud data is then filtered based on the taillight visor structural significance index to remove points with minimal structural impact and retain key structural information points, helping to focus on the key areas of the taillight visor. Using the filtered key 3D point cloud data for matching, the molding quality of the automobile taillight visor can be more accurately detected. By comparing the point cloud data of the actual production taillight visor with the point cloud data of the standard model, deviations and defects in the molding process, such as shape deformation and dimensional errors, can be promptly identified.

[0012] Preferably, the process of obtaining the support degree of the tail cover structure includes:

[0013] Calculate the area of ​​the neighboring point set on the three orthogonal projection planes, then remove the target point cloud from the neighboring point set to obtain a new point set, and calculate the area of ​​the new point set on the three orthogonal projection planes;

[0014] The sum of the area ratio differences between the new point set on the three orthogonal projection planes and the area ratio of the neighboring point set on the three orthogonal projection planes before deleting the target point cloud is calculated as the tail cover structure support of the target point cloud.

[0015] By calculating the area of ​​the neighboring point set on three orthogonal projection planes, the local structural features around the target point cloud can be quantified. After removing the target point cloud, the area of ​​the new point set is calculated and compared with the area of ​​the original neighboring point set to evaluate the impact of the target point cloud on the local structural stability.

[0016] By comparing the change in projected area before and after removing a target point, we can reflect the point's role in maintaining the local geometry. If the area decreases significantly after removal, it indicates that the point is a key support point (such as an edge or corner point); otherwise, it may be a redundant point (such as a point inside a flat area).

[0017] Preferably, the process of obtaining the support degree of the tail cover structure includes:

[0018] The sum of the distances between the target point cloud and all neighboring point clouds in the neighboring point set is calculated to obtain the tail cover structure support of the target point cloud.

[0019] Preferably, the process of obtaining the support degree of the tail cover structure further includes:

[0020] The neighbor aggregation degree of the target point cloud is calculated based on the neighbor point set of the target point cloud, the inverse of the neighbor aggregation degree is used as a weight, and the product of the weight and the tail cover structure support is used as the adjusted tail cover structure support.

[0021] Neighbor clustering measures the degree of clustering of the target point cloud within the neighboring point set. Point cloud regions with high clustering typically contribute more to the stability of the overall structure. By using the inverse of the clustering as a weight, the tail hood structural support can be adjusted to better reflect the importance of key regions.

[0022] Preferably, the process of obtaining the neighbor aggregation degree includes:

[0023] The neighbor point sets of the target point cloud are projected onto three projection planes respectively, and the projection area on each projection plane is calculated. The ratio of the number of neighbor points to the projection area is used as the aggregation component of each projection plane. The aggregation components of the three projection planes are averaged to obtain the neighbor aggregation degree.

[0024] By calculating the nearest neighbor clustering, we can quantify the degree of clustering in a local area of ​​a target point cloud. For example, in point cloud data obtained from 3D scanning, point cloud distribution density and clustering vary across different object surfaces or structures. The nearest neighbor clustering can be used as a characteristic metric to describe this local characteristic, helping to distinguish different point cloud regions, such as flat surfaces, edges, and corners.

[0025] Preferably, the process of obtaining the neighbor aggregation degree includes:

[0026] The neighbor point set of the target point cloud is projected onto three projection planes respectively, and the area and aspect ratio of the minimum circumscribed rectangle on each projection plane are calculated. The average of the ratio of the area of ​​the minimum circumscribed rectangle to the aspect ratio of the minimum circumscribed rectangle of all projection planes is calculated to obtain the neighbor aggregation degree.

[0027] Preferably, the screening process is:

[0028] The tail hood structure significance index of all collected three-dimensional point cloud data is used as input and processed using a clustering algorithm to obtain multiple clusters. The mean of the tail hood structure significance index of all point clouds in each cluster is calculated respectively, and then these means are sorted in ascending order. A set number of clusters are selected and eliminated to obtain the filtered three-dimensional point cloud data.

[0029] By processing three-dimensional point cloud data through a clustering algorithm, similar point cloud data are divided into the same cluster, and then filtered and eliminated according to the mean value of the tail cover structure significance index, a large amount of repeated or similar point cloud information can be removed, thereby reducing data redundancy; and the clustering algorithm can identify outliers or noise points in the data. These points often have a lower tail cover structure significance index. By filtering and eliminating these points, the quality and reliability of the data can be improved.

[0030] Preferably, the clustering algorithm is a DBSCAN clustering algorithm.

[0031] Preferably, the screening process is:

[0032] The three-dimensional point cloud data corresponding to the tail cover structure significance index being less than a preset threshold value is eliminated to obtain filtered three-dimensional point cloud data.

[0033] Preferably, all tail cover structure significance indices are arranged in ascending order and then divided into four equal parts, and the 75th percentile is selected as the threshold for screening.

[0034] The beneficial effects of the present invention are:

[0035] The present invention obtains the tail cover structure support by calculating the area and its changes of the neighboring point set of the point cloud on three orthogonal projection planes, and further considers the adjustment of the tail cover structure support by the neighboring aggregation degree, and comprehensively evaluates the importance of each point cloud in the tail light mask structure from multiple dimensions, and further calculates the tail cover structure significance index and screens the three-dimensional point cloud data based on it. It can effectively identify the areas with important structural features on the tail light mask, exclude irrelevant or interfering point cloud data, and make the subsequent matching detection more accurately focus on the key parts, thereby improving the accuracy of molding quality detection, and being able to more keenly discover the shape deviation, defects and other problems that may occur in the tail light mask during the molding process. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a method flow chart of steps S1 to S2 in a quality inspection method for automobile taillight mask molding according to an embodiment of the present invention.

[0037] Figure 2 It is a schematic diagram of steps S20 to S22 in a quality inspection method for automobile taillight mask molding according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0039] Reference Figure 1 A quality inspection method for automobile taillight mask molding includes steps S1 and S2, which are specifically as follows:

[0040] S1: Collect and preprocess the 3D point cloud data of the car taillight mask.

[0041] In one embodiment, a laser scanner is used to collect three-dimensional point cloud data of a car taillight mask, and each point cloud point contains its coordinate information in space.

[0042] Since factors such as environmental interference during the acquisition process may cause noise, outliers and other noise points in the point cloud, the Gaussian filtering algorithm can be used to denoise the collected three-dimensional point cloud data.

[0043] At the same time, the three-dimensional point cloud data of the automobile taillight mask with standard shape and structure is collected according to the same working environment and steps as the taillight mask template point cloud for subsequent matching with the taillight mask point cloud to be detected.

[0044] S2: Calculate the tail cover structure significance index, filter the 3D point cloud data of the automobile taillight mask based on the calculated tail cover structure significance index, and perform ICP matching between the filtered 3D point cloud data and the 3D point cloud data of the designed shape to evaluate the deviation between the actual shape and the designed shape of the automobile taillight mask, thereby realizing the molding quality inspection of the automobile taillight mask.

[0045] In the point cloud data of a car's taillight mask, the point cloud density varies across different areas of the mask due to the vertical downward scanning direction of the laser scanner. In areas with sparse point clouds, each point cloud point is like an isolated stick in a net, providing relatively strong support to the entire structure because there are no other point cloud points nearby to share the support. For example, the edge of the mask may have fewer point cloud points due to factors such as the scanning angle, making these point cloud points crucial for supporting the mask's edge structure. Conversely, in areas with dense point clouds, the absence of a point cloud point can be compensated by the presence of other surrounding point cloud points, resulting in a smaller impact on the overall structure and a relatively low degree of structural support.

[0046] To quantify this support and assess the significance of the structure corresponding to each point cloud point, a taillight visor structure significance index is constructed. By constructing the taillight visor structure significance index for each point cloud data, key point cloud points that contribute significantly to the structure of the vehicle taillight visor can be identified. These point cloud points are typically located in geometric feature areas such as the ridges and edges of the vehicle taillight visor.

[0047] Reference Figure 2 The acquisition of the above-mentioned tail cover structure significance index includes steps S20 to S22.

[0048] S20: Calculating the tail cover structure support of each three-dimensional point cloud, where the tail cover structure support is used to reflect the degree of change in the tail cover structure after the three-dimensional point cloud is deleted.

[0049] In one embodiment, any one of the three-dimensional point cloud data preprocessed by S1 is selected as the target point cloud, and the K (in the embodiment of the present invention, the value of K is 10, and in other embodiments, the value can be adjusted according to the specific situation) neighboring point clouds closest to the target point cloud are selected to construct the neighboring point set of the target point cloud (the number of point clouds in the neighboring point set of the target point cloud is 11), and the neighboring point set is projected into three orthogonal projection planes XY, XZ, and YZ. For each projection plane, the Alpha Shapes concave hull algorithm is used to connect the outer contour points of the projected plane point cloud to form an area, and the area of ​​the formed area is calculated by the shoelace formula.

[0050] The Alpha Shapes concave hull algorithm can generate a suitable concave polygon based on the point set to describe the shape of the point set, and the shoelace formula is used to calculate the area of ​​the polygon.

[0051] According to the area of ​​each projection plane obtained by the above calculation, the neighbor aggregation degree of the target point cloud is calculated.

[0052] For example, the neighbor aggregation degree of the target point cloud satisfies the following relationship:

[0053]

[0054] Where, is the neighbor aggregation degree of the target point cloud, The neighboring point set of the target point cloud is projected onto the The area behind the projection plane. Indicates the The aggregation component of the projection plane measures the The point cloud density in each projection direction.

[0055] The above method reflects the local density of the target point cloud by calculating the average area of ​​the target point cloud's neighboring point set projected on each projection plane. The larger the value, the wider and more evenly distributed the neighboring point sets of the target point cloud are on the projection plane, that is, the higher the degree of aggregation.

[0056] Exemplarily, another method for calculating the neighbor aggregation degree of the target point cloud is provided, that is, satisfying the relationship:

[0057]

[0058] Where, is the neighbor aggregation degree of the target point cloud, The neighboring point set of the target point cloud is projected onto the The area of ​​the minimum circumscribed rectangle after the projection plane, The neighboring point set of the target point cloud is projected onto the The aspect ratio of the minimum circumscribed rectangle after projection onto the plane.

[0059] The above single projection plane It reflects the specific distribution characteristics of the point set on the plane and is calculated by combining the conditions on all projection planes. Therefore, even if a projection plane The value is large (indicating a narrow and elongated distribution), as long as the distribution on other projection planes is wide and uniform enough, The value can still be large, indicating that the overall clustering of the point set is high.

[0060] Based on the neighbor density of the target point cloud, the taillight cover structure support of the target point cloud is further calculated. The target point cloud is deleted from the neighbor point set to obtain a new point set. The area of ​​the projected area of ​​the new point set is recalculated. The ratio of the projected area of ​​the new point set to the projected area of ​​the neighbor point set before deletion is further calculated to reflect the support degree of the target point cloud for the car taillight cover structure.

[0061] For example, the tail cover structure support of the target point cloud is calculated, that is, the relationship is satisfied:

[0062]

[0063] Where, is the tail cover structure support of the target point cloud, The neighboring point set of the target point cloud is projected onto the The area behind the projection plane, The neighbor point set is projected on the first The area of ​​the projection plane.

[0064] Among them, by comparison and , we can evaluate the impact of the target point cloud on the distribution of neighboring point sets on the projection plane. Indicates the difference in area ratio between the projected area of ​​the neighboring point set before and after deleting the target point cloud. If the projected area before and after the target point cloud is removed changes significantly, then It is larger, indicating that the target point cloud has a greater impact on the structural stability.

[0065] For example, another method for calculating the tail cover structure support of the target point cloud is provided, which satisfies the relationship:

[0066]

[0067] Where, is the tail cover structure support of the target point cloud, Represents the target point cloud and its The Euclidean distance of the nearest neighbor point cloud.

[0068] when When it is larger, it means that the target point cloud is closer to its first The distance between the nearest neighboring point clouds is far, which means that the target point cloud is sparser than its neighboring point clouds in terms of spatial distribution. In the structural analysis of the car taillight mask, if a point is sparser than its neighboring points, it may play a more "independent" or "supporting" role in the structure.

[0069] In the target point cloud, different points contribute differently to the tail hood support due to their position, density, or relationships with other points. By introducing weights, these differences can be quantified, allowing the calculation of tail hood support to more accurately reflect the actual contribution of each point. Weights can be dynamically adjusted based on metrics such as neighbor clustering, allowing for a more refined assessment of tail hood stability. For example, areas with high neighbor clustering have a higher point density and potentially stronger support for the tail hood structure. Therefore, they can be given a smaller weight (because the inverse of the neighbor clustering is smaller, but after multiplying by a larger base support value for the tail hood structure, the overall contribution remains high). Conversely, areas with low neighbor clustering have a lower point density and potentially weaker support for the tail hood structure. Therefore, they are given a larger weight to highlight their relative importance (in the overall calculation, even if the contribution of an individual point is small, the adjusted weight ensures that the overall impact of that area on the tail hood support is not neglected).

[0070] In one embodiment, the inverse of the neighbor aggregation degree of the target point cloud calculated above is used as a weight, and the product of the weight and the tail cover structure support calculated above is used as the adjusted tail cover structure support.

[0071] For example, the tail cover structure support of the adjusted target point cloud is calculated, that is, the relationship is satisfied:

[0072]

[0073] Where, is the tail cover structure support of the adjusted target point cloud, is the tail cover structure support of the target point cloud, It is the inverse of the neighbor aggregation degree of the target point cloud.

[0074] According to the above operation, the tail cover structure support of all collected 3D point cloud data and the adjusted tail cover structure support can be calculated in the same way.

[0075] S21: Calculate the local curvature of each 3D point cloud.

[0076] To further identify the geometric features of the taillight visor, such as ridges, depressions, and protrusions, the local curvature of the point cloud data must be calculated. A least-squares plane algorithm is used to fit a surface to all the point cloud data within the neighborhood of the point cloud data, obtaining a surface equation. This surface equation is then quadratically differentiated, and the resulting second-order derivative is used as the local curvature of the point cloud data. The magnitude of the local curvature reflects the smoothness of the local surface of the point cloud data. Points with large local curvatures are likely to be ridges or edges of the taillight visor, while those with small local curvatures are likely to be located within the surface of the taillight visor.

[0077] S22: Calculate the product of the tail hood structure support and the local curvature of each three-dimensional point cloud in the neighboring point set, and find the maximum and minimum values ​​of the product, take the difference between the product corresponding to the three-dimensional point cloud in the neighboring point set and the minimum value of the product as the first parameter, take the difference between the maximum value of the product in the neighboring point set and the product corresponding to the three-dimensional point cloud in the neighboring point set as the second parameter, and then sum the ratios of the first parameter to the second parameter of each three-dimensional point cloud in the neighboring point set to obtain the tail hood structure significance index of the target point cloud.

[0078] In the taillight hood point cloud to be inspected, a taillight hood structural significance index is calculated by comprehensively considering the taillight hood structural support and local curvature of the point cloud. The taillight hood structural support reflects the degree of support of the point cloud points for the vehicle's taillight hood structure. If a point is located in an important structural area of ​​the taillight hood, such as a ridge or edge, it may make a greater contribution to the overall structure, and therefore its taillight hood structural support will be higher. By considering the taillight hood structural support, those points that are important to the taillight hood structure can be highlighted. The local curvature describes the degree of curvature of the point cloud point location. In the taillight hood, key structural areas such as ridges and edges usually have high curvature variations. By introducing local curvature, these areas with large curvature variations can be captured, thereby identifying the key structures of the taillight hood.

[0079] By calculating the taillight mask structural significance index, we can quantitatively evaluate each point in the taillight mask point cloud data, identifying which points are most likely to be located in key structural areas such as the taillight mask's ridges and edges. These key structural areas are crucial for the design, manufacturing, and quality inspection of taillight masks. By calculating the significance index, we can sort or classify points in the point cloud data according to their structural significance, facilitating subsequent analysis and processing, such as extracting taillight mask features and detecting mask defects. It can be understood that for a three-dimensional point cloud, if the tail hood structure undergoes a significant change after deleting the three-dimensional point cloud, it means that the three-dimensional point cloud can reflect the tail hood structure to a large extent, and the three-dimensional point cloud has a large tail hood structure support; the local curvature can characterize the surface geometric features, and the larger the local curvature, the more complex the geometric features near the three-dimensional point cloud, and the three-dimensional point cloud can also reflect the tail hood structure to a large extent; in order to ensure the accuracy of molding quality inspection, the three-dimensional point cloud that can reflect the tail hood structure should be selected for matching, so the product of the tail hood structure support and the local curvature is used as the tail hood structure significance index to accurately evaluate the ability of the three-dimensional point cloud to reflect the tail hood structure.

[0080] Exemplarily, the significant index of the tail cover structure satisfies the relationship:

[0081]

[0082] Where, is the tail cover structure significance index of the target point cloud, is the first neighbor point in the set The tail cover structure support of a 3D point cloud, For the The local curvature of a 3D point cloud, It represents the minimum value of the product of the tail cover structure support and the local curvature in the neighborhood point concentration. It represents the maximum value of the product of the tail cover structure support and the local curvature in the neighborhood point concentration. is the first parameter, is the second parameter. The 1 in the denominator is a hyperparameter to prevent the denominator from being zero.

[0083] Based on the above operation, the tail cover structure significance index of all collected 3D point cloud data can be calculated. Points with high significance index are more likely to be located in structural feature areas (such as edges or ridges), while points with low significance index are located on smooth surfaces.

[0084] In one embodiment, the tail hood structure significance index of all collected three-dimensional point cloud data is used as input and processed using the DBSCAN clustering algorithm. The clustering algorithm divides the three-dimensional point cloud into different clusters according to the distribution of the tail hood structure significance index of the three-dimensional point cloud.

[0085] The mean of the tail cover structure significance index of all point clouds in each cluster is calculated respectively, and then these means are sorted in ascending order. The first N clusters are selected from the sorting results (N is a hyperparameter, which is 3 in the embodiment of the present invention and can be adjusted according to actual conditions in other embodiments). All point clouds in these N clusters are removed from all collected point cloud data, thereby removing point cloud data with relatively low tail cover structure significance that may interfere with subsequent matching and molding quality detection.

[0086] In another embodiment, a simple, threshold-based direct screening method is provided.

[0087] Exemplarily, the tail hood structure significance indexes of all collected point cloud data are traversed, all tail hood structure significance indexes are arranged in ascending order and then divided into four equal parts, the 75th percentile is selected as the threshold for screening, and then the three-dimensional point cloud data corresponding to the tail hood structure significance index less than the threshold is eliminated to obtain the screened three-dimensional point cloud data.

[0088] After the above screening steps, the filtered 3D point cloud data is obtained. Based on this data, the ICP algorithm is used for matching, and the matching results are used to detect the molding quality of the car taillight cover.

[0089] The ICP algorithm iteratively searches for the optimal transformation relationship between the two sets of point clouds, allowing them to be aligned as closely as possible. This allows the deviation between the actual and designed shapes of the car's taillight mask to be assessed, and the molding quality to be determined.

[0090] It should be noted that those skilled in the art may make various modifications and improvements without departing from the scope of the present invention, and these modifications and improvements fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be based on the appended claims.

Claims

1. A quality inspection method for automobile taillight mask molding, characterized in that: include: Collect 3D point cloud data of car taillight mask and perform preprocessing; Calculate the tail cover structure significance index, filter the 3D point cloud data of the car tail light cover based on the calculated tail cover structure significance index, and perform ICP matching between the filtered 3D point cloud data and the 3D point cloud data of the designed shape to evaluate the deviation between the actual shape of the car tail light cover and the designed shape, thereby realizing the molding quality inspection of the car tail light cover; The calculation process of the tail cover structure significance index is as follows: calculating the tail cover structure support of each 3D point cloud, wherein the tail cover structure support is used to reflect the degree of change of the tail cover structure after deleting the 3D point cloud; calculating the local curvature of each 3D point cloud; Select any three-dimensional point cloud as the target point cloud, determine the neighboring point set of the target point cloud; calculate the product of the tail cover structure support and the local curvature of each three-dimensional point cloud in the neighboring point set, and find the maximum and minimum values ​​of the product, use the difference between the product corresponding to the three-dimensional point cloud in the neighboring point set and the minimum value of the product as the first parameter, use the difference between the maximum value of the product in the neighboring point set and the product corresponding to the three-dimensional point cloud in the neighboring point set as the second parameter, and then sum the ratios of the first parameter and the second parameter of each three-dimensional point cloud in the neighboring point set to obtain the tail cover structure significance index of the target point cloud; The process of obtaining the support degree of the tail cover structure includes: Calculate the area of ​​the neighboring point set on the three orthogonal projection planes, then remove the target point cloud from the neighboring point set to obtain a new point set, and calculate the area of ​​the new point set on the three orthogonal projection planes; Calculate the sum of the area ratio differences between the new point set on the three orthogonal projection planes and the area ratio of the neighboring point set on the three orthogonal projection planes before deleting the target point cloud, and use it as the tail cover structure support of the target point cloud; Alternatively, the sum of the distances between the target point cloud and all neighboring point clouds in the neighboring point set is calculated to obtain the tail cover structure support of the target point cloud.

2. A quality inspection method for automobile taillight mask molding according to claim 1, characterized in that: The process of obtaining the tail cover structure support degree further includes: The neighbor aggregation degree of the target point cloud is calculated based on the neighbor point set of the target point cloud, the inverse of the neighbor aggregation degree is used as a weight, and the product of the weight and the tail cover structure support is used as the adjusted tail cover structure support.

3. The quality inspection method for automobile taillight mask molding according to claim 2, characterized in that: The process of obtaining the neighbor aggregation degree includes: The neighbor point sets of the target point cloud are projected onto three projection planes respectively, and the projection area on each projection plane is calculated. The ratio of the number of neighbor points to the projection area is used as the aggregation component of each projection plane. The aggregation components of the three projection planes are averaged to obtain the neighbor aggregation degree.

4. The method for quality inspection of automobile taillight mask molding according to claim 2, characterized in that: The process of obtaining the neighbor aggregation degree includes: The neighbor point set of the target point cloud is projected onto three projection planes respectively, and the area and aspect ratio of the minimum circumscribed rectangle on each projection plane are calculated. The average of the ratio of the area of ​​the minimum circumscribed rectangle to the aspect ratio of the minimum circumscribed rectangle of all projection planes is calculated to obtain the neighbor aggregation degree.

5. The quality inspection method for automobile taillight mask molding according to claim 3, characterized in that: The screening process is: The tail hood structure significance index of all collected three-dimensional point cloud data is used as input and processed using a clustering algorithm to obtain multiple clusters. The mean of the tail hood structure significance index of all point clouds in each cluster is calculated respectively, and then these means are sorted in ascending order. A set number of clusters are selected and eliminated to obtain the filtered three-dimensional point cloud data.

6. The quality inspection method for automobile taillight mask molding according to claim 5, characterized in that: The clustering algorithm is the DBSCAN clustering algorithm.

7. The quality inspection method for automobile taillight mask molding according to claim 3, characterized in that: The screening process is: The three-dimensional point cloud data corresponding to the tail cover structure significance index being less than a preset threshold value is eliminated to obtain filtered three-dimensional point cloud data.

8. The quality inspection method for automobile taillight mask molding according to claim 7, characterized in that: All tail hood structure significance indices were arranged in ascending order and divided into four equal parts, and the 75th percentile was selected as the screening threshold.

Images