Research On The Method Of Surface Defects Inspection For Copper Strip Based On Machine Vision

Copper strip is a kind of important copper processing materials, which is widely used in electronics, machinery manufacturing and aerospace industries, etc. Surface quality is one of the main factors for evaluating quality of copper strip. Due to rolling equipment, processing technology and environment, copper strip’s surface appears various defects such as scratch, pit and smearing during production process. These defects would seriously affect the quality of copper strip product. Current detection methods have many disadvantages such as low detection efficiency, high inspection missing rate and low accuracy, which can not satisfy the requirements of enterprise. Hence, how to detect the surface defects of copper strip efficiently and accurately is import for improving the quality of copper strip products, which also has important theoretical research value.In this dissertation, a method of surface defects for copper strip is investigated. The main research works are summarized as follows:1. A kind of object detection algorithm is proposed based on local binary patterns. The algorithm can resist the influence of the light better than current methods such as difference image method, coterminous frame difference method, etc. At the same time, the algorithm can also judge whether copper strip images contain surface defects or not better than those methods.2. The traditional image segment algorithm of canny operator is improved based on coterminous frame difference method. The high and low threshold of canny operator can be set adaptively in the algorithm.3. The extraction and selection of characteristics of copper strip surface defects are studied. Copper strip surface defects are expressed in features including grayscale, geometry, shape, and texture. Principal component analysis method is employed to reduce feature dimension and amount of calculation. 4. A classifier of copper strip surface defects based on improved online sequential extreme learning machine is developed. The recognition performance of copper strip surface defects is improved by the classifier.5. A system of copper strip surface defects inspection based on machine vision is developed. The images of copper strip surface can be obtained, transmitted, processed and analyzed automatically. Functions such as defects early warning, storage and query of surface defects, are also implemented.