A 10-micron-level high-precision device assembly surface inspection method

Abstract

The present invention provides a 10-micron-level high-precision device assembly surface detection method, which is characterized by comprising the following steps: Step 1, calculating a device surface estimation model m i ; Step 2, calculate all points in the point cloud data to the estimated model m i distance d; step 3, specify the threshold t, and use the point that satisfies d<t as the intra-office point; step 4, set the estimated model m i Evaluate and get an evaluation score. If the evaluation score is the highest, the current prediction model m i is the best estimation model; step 5, repeat steps 2 to 4, when the number of iterations increases to k, obtain the best model m; step 6, divide the outliers into several point sets based on the clustering segmentation algorithm; step 7 , according to the position of the point set and the positional relationship of the surface to determine the concavity and convexity of the position of the point set, and obtain the maximum surface error of each surface. The invention can accurately and quickly reflect the convexity and concave condition of the assembly surface, and provides convenience for the subsequent device matching determination work.

Description

technical field [0001] The invention belongs to the technical field of virtual assembly, and in particular relates to a method for detecting assembly surfaces of 10-micron-level high-precision devices. Background technique [0002] At present, most virtual assembly systems are based on the assembly constraints of the device, ignoring the fit of the assembly surface, resulting in a large difference between the virtual assembly and the actual assembly process. However, based on the physical properties of the assembly system, dynamic simulation requires a lot of calculations, which is inefficient. [0003] The first thing to do to extract the surface feature information of the assembly object is to segment the data set, that is, to segment the surface point cloud and the non-planar point cloud, and then analyze the unevenness of each surface. At present, there are a large number of surface detection and point cloud segmentation algorithms. According to their working principles...

AI Technical Summary

Problems solved by technology

Although the algorithm can be robustly estimated, when the proportion of inliers in the data to be tested is too small, a large number of iterations is required, which makes the algorithm estimation time increase sharply

Moreover, the random sampling consensus algorithm estimates the quality of the model according to the number of internal points, which often makes the model farther away from the high-density point cloud

Images