A Product Appearance Defect Detection Method Based on Fuzzy Relaxed Constrained Multi-kernel Learning

Abstract

The invention proposes a product appearance defect detection method based on fuzzy relaxation constraint multi-core learning, which belongs to the field of product quality detection of machine vision. Firstly extract some features of real-time collected images; then use fuzzy constraint theory to quantitatively analyze the mapping relationship between features and evaluation indicators; establish a multi-kernel learning model, use multi-kernel learning methods to classify appearance defects, and delineate each kernel function in combination with the mapping quantitative relationship The fuzzy relaxation boundary of the weight; the weight of the multi-core model is determined by the method of fuzzy relaxation constraint (FRC), and the fuzzy range of the weight is determined; finally, the weight size is obtained to obtain the multi-core learning model of different defect types, and the defect detection is obtained by using the multi-core learning model. Test results. The invention adopts the multi-core learning and classification detection method of multi-feature fusion, so that the detection range is wider, combined with fuzzy relaxation constraints to adapt to different detection requirements, can meet the real-time performance of detection, and has high detection accuracy.

Description

technical field [0001] The invention belongs to the field of product quality inspection based on machine vision, and specifically relates to a method for feature extraction and multi-feature information fusion of real-time collected product images and then for product appearance defect detection. Background technique [0002] With the continuous improvement of people's living standards, consumers also have higher quality pursuits for the quality of home appliances, which requires manufacturers to formulate higher production standards for home appliances that can better meet consumer needs. The appearance defect standards of home appliances are shown in Table 1. Area 1 is the bottom and back, Area 2 is the front, and Area 3 is the top and side. New home appliances that leave the factory often have appearance quality problems, which affect the user's consumption experience and cause certain economic losses to the manufacturer. Traditional manual detection methods cannot solve...

AI Technical Summary

Problems solved by technology

[0018] Existing product appearance detection methods involve a wide range of areas, but for large-scale industrial products, such as household appliances, there is no effective method that can meet the real-time detection standards of industrial production. To solve this problem, the present invention provides a method based on fuzzy The product appearance defect detection method of relaxed constraint multi-core learning adopts the detection scheme of multi-feature fusion combined with multi-core learning classification, which can meet the real-time detection and high detection accuracy

Images