A method for denoising measured point cloud of parts based on the change of curve slope

Abstract

The present invention discloses a method for denoising measured point cloud of parts based on curve slope change, which mainly includes the following steps: reading and preprocessing the original point cloud data obtained by measuring the parts; calculating the included angle between adjacent direction vectors of the point cloud curve data points, and reflecting the curve slope change according to the difference between adjacent included angles |θ i -θ i‑1 |; by comparing with the set threshold ||θ||, judging whether P i is a noise point and retaining or removing it; performing multiple iterations to obtain the final point cloud denoising result. The present invention is applicable to the denoising process of the measured point cloud of parts, ensuring that the noise points are effectively removed while ensuring that the valid points in the feature regions of small size and large curvature of the parts are not overly removed, effectively improving the accuracy and efficiency of removing the noise points in the point cloud.

Description

Technical Field

[0001] The present invention relates to the technical field of point cloud denoising, and particularly relates to a method for denoising measured point clouds of parts based on the change of curve slope. Background Art

[0002] In recent years, robot intelligent processing technology has been gradually applied to the processing of parts in the fields of aviation, aerospace, weapons, etc. However, the shapes of aviation parts are complex and their structures are diverse, and traditional measurement methods are difficult to meet the measurement requirements for robot processing trajectory planning. The rapid development of three-dimensional data acquisition technology has made it possible to directly obtain high-density point cloud data on the surface of parts, and the acquisition and processing of point cloud data play a crucial role in robot processing trajectory planning.

[0003] Affected by various factors such as the measured part itself, the electrical characteristics of the sensor, and the measurement environment, the acquired point cloud data often contains a certain number of noise points. At the same time, aviation parts often have higher precision requirements. For example, the leading and trailing edges of aviation engine blades are very thin, that is, the radius of the leading and trailing edge arcs is very small and the curvature is large. Using traditional data filtering methods may filter out the large curvature part as noise, and after adjusting the filtering effect, some noises similar to the curvature characteristics of the leading and trailing edges cannot be filtered out, which cannot meet the requirements of high-efficiency and high-precision point cloud denoising in the fields of aviation and aerospace. To solve the above technical problems, the present invention proposes a curve inspection method based on slope change for point cloud denoising processing to achieve the denoising processing of measured point cloud data of parts. Summary of the Invention

[0004] The technical problems to be solved by the present invention are as follows:

[0005] To meet the intelligent processing requirements of parts, it is necessary to achieve efficient and high-precision point cloud acquisition and processing in small-size, large-curvature and other feature regions. Using traditional filtering methods, there is a problem that noise points are difficult to be correctly removed. On the premise of ensuring processing accuracy and improving processing efficiency, how to ensure that noise points are effectively removed while valid points are not over-removed is crucial for the point cloud denoising process of parts.

[0006] Based on the acquisition and preprocessing of three-dimensional point cloud data, the present invention uses a curve inspection method based on slope change to filter noise points from point cloud slice data, realizes the denoising processing of measured point cloud data of parts, and improves the point cloud denoising efficiency.

[0007] The technical solution adopted by the present invention is: a method for denoising measured point clouds of parts based on the change of curve slope, and the method flow is as follows:

[0008] Step 1: Read the original measured point cloud data of the part into the memory and perform slicing and sorting processing to obtain the point cloud data to be denoised.

[0009] Step 2, calculate the angle θ between adjacent direction vectors for the i-th point cloud curve data point P i specifically, the calculation method is to calculate the direction vectors of adjacent points i and the direction vector to obtain the angle θ between them. Calculate the direction vectors of adjacent points i-1 and the direction vector to obtain the angle θ between them. The obtained difference between adjacent angles |θ i -θ i -θ i-1 | reflects the change in the curve slope, that is, the slope change angle.

[0010] Step 3, determine whether the slope change angle |θ i -θ i-1 | exceeds the set threshold ||θ||. If it does not exceed, retain the P i point. If |θ i -θ i-1 | exceeds the set threshold ||θ|| and θ i also exceeds ||θ||, then determine that Pi is a noise point and remove this point.

[0011] Step 4, repeat Steps 2 - 3 for multiple iterations to obtain the final point cloud denoising result.

[0012] Furthermore, the point cloud slicing method in Step 1 is as follows: select the plane E to be sliced, set the slice thickness 2δ according to the sampling density of the measured point cloud data of the part, and generate planes E1 and E2 at a distance of δ on both sides of the plane E. Use the data bounding box of the octree Octree to obtain all the data points located between the planes E1 and E2, project these points onto the plane E, and fit all the projected points to obtain the intersection line of the cutting plane E and the point cloud data.

[0013] Furthermore, the point cloud sorting method in Step 1 is as follows: consider the cross-sectional profile characteristics of the part to sort the unordered point set obtained after slicing, and initially obtain the point cloud data slice curve formed by connecting each data point.

[0014] Furthermore, the method for selecting the threshold ||θ|| in Step 3 is as follows: calculate the standard deviation σ of the curve slope change angle |θ i -θ i-1 |, and determine the threshold ||θ|| with this as a reference value.

[0015] Furthermore, Step 4 is to traverse all the points in the point cloud dataset, judge whether each point is a noise point one by one, and remove all the noise points to obtain the preliminarily filtered point set. Brief Description of the Drawings

[0016] Figure 1Flow chart of the method of the present invention

[0017] Figure 2 Schematic diagram of the curve inspection method based on the slope change angle

[0018] Figure 3 Point cloud data curve before noise point removal processing

[0019] Figure 4 Slope change angle of the curve before noise point removal processing

[0020] Figure 5 Slope change angle of the curve after noise point removal processing

[0021] Figure 6 Point cloud data curve after noise point removal processing Specific embodiments

[0022] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. The following embodiments do not limit the present invention.

[0023] This method can be applied to the processing of the measured point cloud of each part. When specifically implemented, refer to Figure 1 , read in the three-dimensional point cloud data obtained from the actual measurement of the part, and perform point cloud denoising processing after slicing and sorting. In this embodiment, the single-layer grinding processing path curve of the aero-engine blade is used as the implementation object for point cloud denoising processing.

[0024] The technical solution adopted by the present invention is: a method for denoising the measured point cloud of a part based on the slope change of the curve, and the method flow is as follows:

[0025] Step 1: Read the original measured point cloud data of the part into the memory and perform slicing and sorting processing to obtain the point cloud data to be denoised.

[0026] Step 2: Calculate the angle θ i between the adjacent direction vectors of the i-th point cloud curve data point P i specifically, the calculation method is to calculate the direction vectors of adjacent points and the direction vector of the angle θ i-1 calculate the direction vectors of adjacent points and the direction vector of the angle θ i , and obtain the adjacent angle difference |θ i -θ i-1 | to reflect the slope change of the curve, that is, the slope change angle.

[0027] Step 3: Judge whether the slope change angle |θ i -θ i-1 | exceeds the set threshold ||θ||, if not, then retain Pi At a point, if |θ i - θ i-1 | exceeds the set threshold ||θ|| and θ i also exceeds ||θ||, then it is determined that P i is a noise point and this point is removed.

[0028] Step 4: Repeat Steps 2 - 3 for multiple iterations to obtain the final point cloud denoising result.

[0029] Furthermore, the point cloud slicing method in Step 1 is as follows: Select the plane E to be sliced, set the slice thickness 2δ according to the sampling density of the point cloud data, and generate planes E1 and E2 at a distance of δ on both sides of the plane E. Use the data bounding box of the octree Octree to obtain all the data points located between the planes E1 and E2, project these points onto the plane E, and the fitted line of all the projected points can be used to obtain the intersection line of the cutting plane E and the point cloud data.

[0030] Furthermore, the point cloud sorting method in Step 1 is as follows: Consider the part cross - section profile characteristics to sort the unordered point set obtained after slicing, and initially obtain the point cloud data slice curve formed by connecting each data point. For the blade part, the point cloud data to be denoised obtained after processing the measured point cloud of the part through Step 1 is as Figure 3 shown.

[0031] Furthermore, the method for selecting the threshold ||θ|| in Step 3 is as follows: Calculate the standard deviation σ of the change angle |θ i - θ i-1 | of the curve slope, and determine the threshold ||θ|| with this as a reference value.

[0032] Furthermore, Step 3 is to traverse all the points in the point cloud dataset, judge one by one whether they are noise points and remove all the noise points to obtain a preliminarily filtered point set.

[0033] The change angles of the slopes of the curves before and after the point cloud denoising process are as Figure 4 and Figure 5 shown. It can be seen that after the processing, the fluctuation of the change angle of the slope is significantly reduced, indicating that the processed curve is smoother. By comparing the single - layer grinding processing path curves of the blade before and after the denoising process, as Figure 2 and Figure 6 shown, it can be seen that the number of noise points processed by this method is significantly reduced, and the structural characteristics at the front and rear edges of the part remain unchanged.

[0034] The present invention adopts a method for denoising the measured point cloud of a part based on the change of the curve slope. The change angle of the slope on the point cloud curve is calculated using the point cloud data slicing result, and compared with the set threshold to determine whether each data point is a noise point, thereby realizing the denoising of the point cloud data.

[0035] The advantages of the present invention compared with the prior art are as follows: for the characteristic regions with small part sizes and large curvatures, the traditional filtering method has the problem that it is difficult to correctly remove noise points. The present invention proposes a curve inspection method based on slope changes, which can realize the denoising process of the measured point cloud data of parts, ensure that valid points are not over-removed, has a fast processing speed and high efficiency, effectively solves the problems of low accuracy and low efficiency of the traditional method, and thus ensures the efficiency and accuracy of the subsequent analysis and processing of the measured point cloud data of parts.

Claims

1. A method for denoising measured point cloud of parts based on curve slope change, and its method process is as follows: Step 1: Read the original measured point cloud data of the part into memory and perform slicing and sorting processing to obtain the point cloud data to be denoised. Step 2: Calculate the i-th point cloud curve data point P i successively, and calculate the included angle θ i between adjacent direction vectors. The specific calculation method is to calculate the direction vectors of adjacent points and the direction vector to obtain the included angle θ i-1 . Calculate the direction vectors of adjacent points and the direction vector to obtain the included angle θ i . The obtained difference between adjacent included angles |θ i -θ i-1 | reflects the change in the curve slope, that is, the slope change angle. Step 3, determine whether the slope change angle |θ i -θ i-1 | exceeds the set threshold ||θ||. If it does not exceed, retain point P i ; if |θ i -θ i-1 | exceeds the set threshold ||θ|| and θ i also exceeds ||θ||, then determine that P i is a noise point and remove this point Step 4: Repeat Steps 2 - 3 for multiple iterations to obtain the final point cloud denoising result.

2. The method for denoising measured point cloud of parts based on the change of curve slope according to claim 1, wherein, The point cloud data slicing method in Step 1 is as follows: Use the cutting plane E to slice the point cloud data, set the slice thickness 2δ according to the sampling density of the point cloud data, project the points at a distance of ±δ from the plane E onto the plane E, and fit all the projected points to obtain the intersection line of the cutting plane E and the point cloud data.

3. A method for denoising measured point clouds of parts based on the change of curve slope according to claim 1, characterized in that The point cloud sorting method in Step 1 is as follows: Consider the cross-sectional profile characteristics of the part to sort the unordered point set obtained after slicing, so as to obtain the point cloud data slicing curve formed by connecting each data point.

4. A method for denoising measured point cloud of parts based on curve slope change according to claim 1, characterized in that, The selection method of the threshold ||θ|| in step 3 is to calculate the standard deviation σ of the curve slope change angle |θ i -θ i-1 | and determine the threshold ||θ|| based on this reference value.

Images