Real-time defect detection based on symmetric difference and component match for high-speed moving yarn

SHEN Ling-fei， MENG Xu-jiong
（Institute of Intelligence and Software Technology， Hangzhou Dianzi University， Hangzhou 310018， China）

Abstract： Aiming at the problems of a real-time defect detection algorithm for high-speed yarn，this algorithm was used to solve the defect problem of real-time to accurately detecting. The algorithm was suitable for real-time processing of large-volume image and high-speed moving yarn defect detection. By the symmetrical differencing algorithm and connected component matching，the algorithm was combined to improve the yarn defect detection accuracy while reducing processing time. Firstly，image needed preprocessing，then images were decomposed and extracted by using symmetrical differencing algorithm，and then，defects were recognized by using the connected component. The shortcoming of anti-tremble of the traditional differential algorithm was improved. Component matching method with particularly strong ability of identifying was constructed. Finally，the algorithm's defect detection accuracy and detection speed were compared with existing detection methods analysis. The results indicate that this algorithm is better than manual detection and traditional algorithms，detection speed relative to the neural network and the traditional algorithms has been improved. It is considered that the algorithm is able to achieve real-time rapid detection of defects at the same time ensure the detection accuracy of defect detection.
Key words： defect detection； real-time processing； high-speed moving yarn； symmetrical differencing algorithm； connected component