Yarn Anomaly Detection Using K-th Moment and Entropy Filters

An optical yarn quality measurement system is constructed. To extract yarn defects of neps and foreign fibers with different colors, a novel structure of optoelectronic sensor is proposed. The experimental signals acquired by the system perform low signal to noise ratio therefore makes it difficult to defect discrimination. Concerning these problems, two methods, K-th moment filter and minimum entropy filter, are utilized for signal processing. The results indicate that both methods are effective for the extracting of abnormal signals corresponding to yarn defects. The K-th moment filter has the advantage of lower computational complexity and is more effective in on-line measurement system. On the other hand, the minimum entropy filter can remain more local tiny features in the processed signals by adjusting the parameters properly so that is better in anomaly detecting of the defects in different levels.