A method for outlier removal in point cloud processing of soldering boards

Abstract

This invention discloses an outlier removal method for solder plate point cloud processing. First, an image of the solder plate point cloud to be processed is acquired. The solder plate plane is segmented, and equations for each plane are fitted. Then, the positions of the theoretical intersection lines of the solder plates are determined based on the equations. Based on the positions of the theoretical intersection lines and the coordinates of each point in the point cloud image, the solder plate plane to which each point belongs is determined. Finally, the distance from each point to its corresponding solder plate plane is determined, and a distance threshold and processing range are set. If the distance is less than the threshold, the point is retained; otherwise, it is identified as an outlier and deleted. This invention incorporates the shape information of the solder plate plane during processing, making the denoising process simpler and faster. It can remove not only generalized outliers but also has good removal effects on other types of outlier noise points.

Description

Technical Field

[0001] This invention relates to the field of solder plate point cloud repair technology, and in particular to a method for removing outliers in solder plate point cloud processing. Background Technology

[0002] Point clouds are massive collections of points describing the surface characteristics of an object, typically obtained using LiDAR or scanners. When acquiring point clouds of solder plates using a contour scanner, due to issues such as partial corrosion and metallic reflection on the solder plate surface, the accuracy of the visual sensor, and potential occlusion, some points in the scan results may shift in position, forming outliers and holes. Outliers are a type of noise and redundant parts of the point cloud, which should be removed as they negatively impact subsequent reconstruction and feature recognition processes. Existing technologies mostly use smoothing filters for outlier removal; however, these filters can only remove outliers in a general sense. When a particular outlier is surrounded by many other points, it becomes difficult to remove, resulting in poor processing performance. Summary of the Invention

[0003] The purpose of this invention is to address the shortcomings of existing technologies by proposing an outlier removal method for soldering board point cloud processing. Compared to smoothing filters, which can only remove outliers in a general sense, this invention has a good removal effect on various types of outlier noise points.

[0004] The technical solution to achieve the objective of this invention is:

[0005] A method for outlier removal in point cloud processing of solder pads includes the following steps:

[0006] Step S1: Obtain the point cloud image of the soldering board to be processed;

[0007] Step S2: Divide the solder plate plane and fit the solder plate plane equations respectively;

[0008] Step S3: Determine the position of the theoretical intersection line of the solder plate according to the solder plate plane equation, and determine the solder plate plane to which each point belongs based on the position of the theoretical intersection line and the coordinates of each point in the solder plate point cloud image.

[0009] Step S4: Determine the distance from each point to its corresponding soldering plate plane, set a distance threshold and processing range. If the distance is less than the threshold, retain the point; otherwise, identify it as an outlier and delete it.

[0010] Furthermore, in step S1, a contour scanner is installed on a moving platform, and the moving platform is controlled to move at a constant speed along the weld seam, thereby driving the sensor of the contour scanner to scan and acquire data, and convert it into an image of the weld plate point cloud.

[0011] Furthermore, by substituting the laser length l of the contour scanner, the number of data columns n, the sampling frequency f of the contour scanner, and the movement speed v of the moving platform into the following formula, the actual distance between the x and y coordinates of two points in the solder plate point cloud is calibrated, and the data acquired by the contour scanner's sensor is converted into an image of the solder plate point cloud.

[0012]

[0013] y_step = v·f

[0014] Where x_step and y_step represent the actual distance between the two points in terms of their x and y coordinates.

[0015] Further, in step S2, the three-dimensional image data of the weld plate point cloud is reduced to one-dimensional information by principal component analysis, key features are extracted and feature values ​​of the weld plate point cloud are obtained, and two weld plate planes are segmented according to the difference between the weld seam and the feature values ​​of the weld plate point cloud. The least squares method is used to fit the equations ax + by + cz + d = 0 of the weld plate planes, where a, b, c, and d are the four parameters that make up the weld plate planes.

[0016] Furthermore, the dimensionality reduction of the point cloud on the welding plate specifically includes the following steps:

[0017] Step S21: Select the number of segments according to the accuracy of the weld, the number of segments is in the range of 5 to 10, and divide the entire weld plate point cloud into segments along the x direction;

[0018] Step S22: Determine the feature vector E and the corresponding feature value Λ for each segment of the solder plate point cloud, and obtain the proportion of the maximum feature value for each segment;

[0019] Step S23: Suppose there are m n′-dimensional data before dimensionality reduction. Take the original data n′ rows and m columns matrix X, and arrange the eigenvectors E in rows from top to bottom according to the size of the corresponding eigenvalues ​​Λ to form matrix N. Take the first j rows of matrix N to form matrix P, and you can obtain the dimensionality-reduced data Y = PX, where j is the dimension to be reduced.

[0020] Further, in step S3, the entire weld plate point cloud is divided into k segments along the weld direction, each segment having a length of len. Then, the value of yi at the theoretical feature point in the i-th segment is...

[0021]

[0022] Thus, the position of the theoretical intersection line is obtained, and the plane equations of the two divided welding plates are established as follows:

[0023] a1x + b1y + c1z + d1 = 0

[0024] a²x + b²y + c²z + d² = 0

[0025] Where a1, b1, c1, d1 and a2, b2, c2, d2 are four parameters of the two welding plate planes, thus determining the values ​​of xi and zi for the theoretical feature points in the i-th segment.

[0026]

[0027]

[0028] Based on the position of the weld plate dividing line obtained from the theoretical intersection line and the coordinates of each point, determine the weld plate plane to which each point belongs.

[0029] Further, in step S4, the distance threshold is twice the weld width, and the processing range is the area of ​​the weld plate point cloud excluding points near the theoretical intersection line.

[0030] By adopting the above technical solution, the present invention has the following beneficial effects:

[0031] (1) The present invention incorporates the shape information of the solder plate plane during the processing, making the noise reduction process simpler and faster. It can not only remove outliers in a general sense, but also has a good removal effect on other types of outlier noise points.

[0032] (2) The present invention uses a mobile platform to drive the contour scanner to scan at a constant speed, ensuring that the acquired data is more accurate.

[0033] (3) Compared with the traditional outlier removal method, this invention abandons the computationally intensive region growing algorithm in the solder plate segmentation stage and adopts the principal component analysis method with extremely fast computation speed, which greatly reduces the processing time. By reducing the dimensionality of three-dimensional image data to one-dimensional information, the amount of data processing is greatly reduced.

[0034] (4) By pre-setting the processing range, the present invention effectively prevents the points at the weld from being accidentally deleted, thereby improving the accuracy of data processing. Attached Figure Description

[0035] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein:

[0036] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0037] To better understand the above technical solutions, a detailed description of the solutions will be provided below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0038] (Example 1)

[0039] like Figure 1 The outlier removal method shown for solder pad point cloud processing includes the following steps:

[0040] Step S1: Obtain the point cloud image of the soldering board to be processed;

[0041] First, a contour scanner is mounted on a moving platform. The platform is then controlled to move uniformly along the weld seam at a preset speed, thereby driving the scanner's sensor to scan and acquire data. The moving platform ensures more accurate data acquisition. Since the data type read directly from the sensor is not a point cloud data type, the data format needs to be converted first. This involves calibrating the actual distance between two points in the weld plate point cloud using the x and y coordinates. This is done by substituting the laser length *l* of the contour scanner, the number of data columns *n*, the scanner's sampling frequency *f*, and the moving platform's speed *v* into the following formula. This process converts the data acquired by the contour scanner's sensor into an image of the weld plate point cloud.

[0042]

[0043] y_step = v·f

[0044] Where x_step and y_step represent the actual distance between the two points in terms of their x and y coordinates.

[0045] Step S2: Divide the solder plate plane and fit the solder plate plane equations respectively;

[0046] The three-dimensional image data of the obtained solder plate point cloud is reduced to one-dimensional information using principal component analysis, specifically including the following steps:

[0047] Step S21: Select the number of segments according to the accuracy of the weld. Divide the entire weld plate point cloud into segments along the x direction. Too many segments will affect the processing time, while too few segments will result in a large error. The range of the number of segments is 5 to 10. In this example, the preferred number of segments is 7.

[0048] Step S22: Determine the feature vector E and the corresponding feature value Λ for each segment of the solder plate point cloud, and obtain the proportion of the maximum feature value for each segment. Use the proportion of the maximum feature value instead of the maximum feature value directly for subsequent processing to prevent the difference in the number of points in different segments from affecting the results.

[0049] Suppose there are m n′-dimensional data points before dimensionality reduction. Transform the original data into an n′-row, m-column matrix X.

[0050] X = (a1 a2 … a M )

[0051] Each row of X represents an attribute. For each row, we remove the zero-mean, i.e., subtract the average value of that row, to obtain...

[0052]

[0053] The covariance matrix C is

[0054]

[0055] The eigenvectors E and corresponding eigenvalues ​​Λ of the covariance matrix C are

[0056] E = (e1 e2 … e) n )

[0057]

[0058] Step S23: Arrange the eigenvectors E into matrix N by row from top to bottom according to the size of the corresponding eigenvalues ​​Λ. Take the first j rows of matrix N to form matrix P, and you can obtain the dimension-reduced data Y = PX, where j is the dimension to be reduced.

[0059] Similarly, to prevent the number of points on the outermost two sides of the point cloud from affecting the results, the leftmost and rightmost points of the entire weld plate point cloud are excluded. In this embodiment, the number of segments is 7, that is, the 1st and 7th segments are excluded. The part with the smallest proportion of the largest eigenvalue is taken. The center of this part is the location of the weld. The entire weld plate point cloud is divided into two weld plate planes by this boundary line. The least squares method is used to fit the equations of the weld plate planes ax + by + cz + d = 0, where a, b, c, and d are the four parameters that make up the weld plate planes.

[0060] Step S3: Determine the position of the theoretical intersection line of the solder plate according to the solder plate plane equation, and determine the solder plate plane to which each point belongs based on the position of the theoretical intersection line and the coordinates of each point in the solder plate point cloud image.

[0061] First, divide the entire weld plate point cloud into k segments along the weld direction, each segment having a length of len. In this preferred example, len is 3. Then, the value of yi in the theoretical feature point (xi, yi, zi) of the i-th segment is...

[0062]

[0063] The plane equations of the two split welding plates are established as follows:

[0064] a1x + b1y + c1z + d1 = 0

[0065] a²x + b²y + c²z + d² = 0

[0066] Where a1, b1, c1, d1 and a2, b2, c2, d2 are four parameters of the two welding plate planes, thus determining the values ​​of xi and zi for the theoretical feature points in the i-th segment.

[0067]

[0068]

[0069] Based on the position of the weld plate dividing line obtained from the theoretical intersection line and the coordinates of each point, determine the weld plate plane to which each point belongs.

[0070] Step S4: Traverse the point cloud and determine the distance from each point to its corresponding weld plate plane. Set a distance threshold based on the weld width, which is twice the weld width. In this embodiment, the weld width is 0.5mm, and the distance threshold is set to 1mm. To prevent the deletion of points at the weld, points near the theoretical intersection line are first excluded. Based on the x-coordinate values ​​of each point, a pass-through filter is used to filter out points whose x-coordinate values ​​meet the following conditions.

[0071] x i ∈(x bound -thre,x bound +thre)

[0072] Where x boun d represents the x-coordinate value at the theoretical intersection line, and thre represents the threshold for excluding welds, thus setting a processing range.

[0073] Fit the plane equation of one of the welding plates as follows

[0074] ax + by + cz + d = 0

[0075] The distance from each point to its corresponding welding plate plane is

[0076]

[0077] Where x0, y0, and z0 are the coordinates of each point. If the distance dis is less than the distance threshold of 1 mm, the point is retained; otherwise, it is identified as an outlier and deleted.

[0078] This invention incorporates the shape information of the solder plate plane during processing, making the denoising process simpler and faster. It not only removes generalized outliers but also demonstrates excellent removal performance for other types of outliers. Compared to traditional outlier removal methods, the computationally intensive region growing algorithm is abandoned in the solder plate segmentation stage, replaced by the extremely fast principal component analysis method, significantly reducing processing time. By reducing the dimensionality of three-dimensional image data to one-dimensional information, the amount of data processed is greatly reduced.

[0079] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for removing outliers in point cloud processing of soldering boards, characterized in that, Includes the following steps: Step S1: Obtain the point cloud image of the soldering board to be processed; Step S2: Segment the weld plate plane and fit the weld plate plane equations separately. Principal component analysis is used to reduce the dimensionality of the obtained 3D image data of the weld plate point cloud to 1D information. Key features are extracted and feature values ​​of the weld plate point cloud are obtained. Based on the difference between the weld seam and the feature values ​​of the weld plate point cloud, two weld plate planes are segmented. The least squares method is used to fit the weld plate plane equation ax + by + cz + d = 0, where a, b, c, and d are the four parameters that make up the weld plate plane. The dimensionality reduction of the weld plate point cloud specifically includes the following steps: Step S21: Select the number of segments according to the weld width and divide the entire weld plate point cloud into segments along the x direction; Step S22: Determine the feature vector E and the corresponding feature value Λ for each segment of the solder plate point cloud, and obtain the proportion of the maximum feature value for each segment; Step S23: Given m n′-dimensional data, the original data is an n′-row m-column matrix X. The eigenvectors E are arranged from top to bottom into matrix N according to the size of their corresponding eigenvalues ​​Λ. The first j rows of matrix N are taken to form matrix P, and the dimension-reduced data Y = PX can be obtained. Step S3: Determine the position of the theoretical intersection line of the solder plate according to the solder plate plane equation, and determine the solder plate plane to which each point belongs based on the position of the theoretical intersection line and the coordinates of each point in the solder plate point cloud image. Step S4: Determine the distance from each point to its corresponding soldering plate plane, set a distance threshold and processing range. If the distance is less than the threshold, retain the point; otherwise, identify it as an outlier and delete it.

2. The outlier removal method for point cloud processing of soldering boards according to claim 1, characterized in that: In step S1, a contour scanner is installed on a moving platform, and the moving platform is controlled to move at a constant speed along the weld seam, thereby driving the sensor of the contour scanner to scan and acquire data, and convert it into an image of the weld plate point cloud.

3. The outlier removal method for point cloud processing of soldering boards according to claim 2, characterized in that: By substituting the laser length *l* of the contour scanner, the number of data columns *n*, the sampling frequency *f* of the contour scanner, and the movement speed *v* of the moving platform into the following formula, the actual distance between the x and y coordinates of two points in the solder plate point cloud is calibrated, and the data acquired by the contour scanner's sensor is converted into an image of the solder plate point cloud. y_step = v·f Where x_step and y_step represent the actual distance between the two points in terms of their x and y coordinates.

4. The outlier removal method for point cloud processing of soldering boards according to claim 1, characterized in that: In step S3, the entire weld plate point cloud is divided into k segments along the weld direction, each segment having a length of len. The value of yi at the theoretical feature point in the i-th segment is then determined. Thus, the position of the theoretical intersection line is obtained, and the plane equations of the two divided welding plates are established as follows: a1x + b1y + c1z + d1 = 0 a²x + b²y + c²z + d² = 0 Thus, the theoretical feature point xi and zi value in the i-th segment are determined. Based on the position of the weld plate dividing line obtained from the theoretical intersection line and the coordinates of each point, determine the weld plate plane to which each point belongs.

5. The outlier removal method for point cloud processing of soldering boards according to claim 1, characterized in that: In step S4, the distance threshold is twice the weld width, and the processing range is the area where the weld plate point cloud is located, excluding points near the theoretical intersection line.

Images