Dissertation > Industrial Technology > Automation technology,computer technology > Automation technology and equipment > Automation systems > Data processing, data processing system > Centralized testing and roving detection system

Research on Surface Defects Detection for Steel Strips Based on Machine Vision

Author JiaFangQing
Tutor YangTianZuo
School Chongqing University
Course Pattern Recognition and Intelligent Systems
Keywords Defect detection Neural network Rolled strips Machine vision
CLC TP274.4
Type Master's thesis
Year 2007
Downloads 407
Quotes 11
Download Dissertation

Quality control has gained high significance in steel production. Defects on the surface of steel strips are main factors to evaluate quality of steel strips, so on-line surface inspection is of great importance to improve quality of steel strips. Traditional visual surface inspection by human inspectors is far from satisfactory. Developing automatic surface inspection systems is a common demand of steel corporations.Surface defects inspection techniques based on Machine Vision are investigated. The general design scheme of surface defects inspection system for cold rolled strips is given, hardware and software of the system is described, and algorithms for detect detection、image feature extraction and defect classification are discussed. The details are as follows:①Based on the technology requirement of surface defects inspection system for cold rolled strips, the article gives the general design scheme of the system. Multiple CCD matrix cameras are equipped to capture surface images of steel strips simultaneously, which leads to high precision of inspection. Parallel computer system consisting of multiple clients and a server offers possibility for on-line inspection from hardware.②Algorithms for defect detection are developed. Image processing, such as filter smoothing the noise and image edge detection applied in surface detection are studied and developed, according to the emphasis on precision requirement, which were more suits for the features extraction.③Features extracted by the system are discussed. A feature selection method based on back-propagation networks is proposed. The combine of image feature extraction and defect classification based on back-propagation networks is more valid than other method.④Algorithms for defect classification are developed. Classifiers are constructed based on back-propagation network. Experiments with samples of surface detects show that classification rate is up to 89%.

Related Dissertations
More Dissertations