Surface defect detection method

Abstract

The surface defect detection method comprises the following steps: collecting an image of a defect through a camera; the method comprises the following steps: firstly, collecting a surface defect, then segmenting the collected image, respectively inputting segmented sub-images into a DCNN network and an Adaboost network, respectively outputting DCNN features and Adaboost features through the DCNN network and the Adaboost network, finally, carrying out normalization fusion on the features, classifying the features by adopting a classifier, outputting the type of the surface defect and the probability of belonging to the type, and finally, outputting the probability of belonging to the type of the surface defect. According to the method, detection and classification are carried out by adopting an Adaboost and DCNN fusion mode under the condition that defect image training samples are not enough, so that the recognition precision of the surface defects is remarkably improved, and a basis is provided for detecting the surface defects of complex irregular objects.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a surface defect detection method. Background technique [0002] Machine vision technology has replaced human eyes and penetrated into all aspects of society, completely changing people's living environment. Machine vision detection combines machine vision and automation technology, and is widely used in product defect detection in the manufacturing industry, such as product assembly process detection and positioning, product packaging detection, product appearance quality detection, cargo sorting or fruit sorting in the logistics industry Picking, etc., machine vision can replace manual work to complete various tasks quickly and accurately. [0003] The present invention mainly aims at the problem that the detection of surface defects such as large-scale workpiece surface crack detection, aircraft skin defect detection, and screw corrosion surface detection is not easy...

AI Technical Summary

Problems solved by technology

[0004] Application No. 201910264717.X uses image preprocessing and PixelNet network to segment defect images, but does not perform defect recognition on defect surfaces; Application No. 201810820348.3 introduces attention module into convolution module to improve detection accuracy, but increases training difficulty, so , the present invention mainly proposes a surface defect detection method in the case of insufficient crack image training samples, which can effectively improve the recognition accuracy of surface defects and provide a basis for the detection of surface cracks on complex and irregular objects

Images