Detection of Residual Yarn in Bobbin Based on Hough Transform and BP Neural Network

In order to promote the intellectualization of textile industry, in this paper, a step-by-step residual yarn detection algorithm based on Hough transform and BP neural network is proposed. According to the geometric characteristic's transformation of bobbins with a large amount of residual yarn, the images are converted to the grayscale, then the edges of image are detected by Canny-Otsu algorithm and the outlines of bobbin are detected by Hough transform, the bobbins with a large amount of residual yarn are eliminated by outline detection results. The images which cannot eliminated by outline detection are converted to HSV color space and divide to several sub regions, the HSV values are extracted as the input layer of BP neural network, the residual yarn condition of bobbins as the output layer. The experimental results show that the detection accuracy of algorithm is up to 96%, which meets the requirements of practical application.