Interval Clustering Method for Component Defect Detection Based on Feature Selection and Combination Optimization Algorithm

Abstract

The invention discloses a component defect detection interval clustering method based on feature selection and combined optimization algorithm. The specific steps are as follows: the first step, collecting data; the second step, data cleaning; the third step, balancing data distribution; the fourth step Step, feature selection; step 5, setting interval combination, selecting a better interval combination and performing expansion optimization on it; step 6, removing the data in the combination from the original data set, and repeating step 5 for the remaining data until all positive The samples are all selected by the rules, and a series of rule descriptions are obtained, and the combinatorial optimization approaches the end of the algorithm. This method performs combined optimization and clustering of positive and negative samples on each optical surface of different defects of industrial parts, and has certain robustness to ensure accurate detection and classification of multi-item defects.

Description

technical field [0001] The invention relates to the technical field of image data processing, in particular to a component defect detection interval clustering method based on a feature selection and combination optimization algorithm. Background technique [0002] At present, most methods based on image data processing select physical quantity intervals for clustering based on experience. Differences in physical quantity weights, optical surfaces, and defect types affect the accuracy of positive and negative sample division, and there are many limitations. The most obvious is that the weight of the length and width physical quantities of linear defects is relatively large, and the physical quantity of area is not considered; for massive defects, the weight of the physical quantity of area is relatively large, and the physical quantity of length and width is not considered. This leads to suboptimal results for some interval combinations. At the same time, the optical surfac...

AI Technical Summary

Problems solved by technology

This leads to some interval combinations are not a better result

At the same time, the optical surface of the same defect is different, which makes the combination of setting intervals complicated

Images