The Research Of Cord Fabric Defects Detection Based On Image Processing

Textile defects occur in the process of weaving, and the defects impact the grade of cord fabric quality, and the sales and exports. The purpose of this paper is to find a rapid approach of the cord fabric defects detection based on computer vision.Based on the study and research of the development of the automatic detecting technology, this paper offers a method of the cord fabric defects detection. In the process of rapid defects detection, this paper chose the method based on image entropy to resolve the problem of rapid approach of the cord fabric defects detection. This paper provides entropy matrix of the sub-image segmented from the original image, and detects whether the image has defects by the difference of the maximum and the minimum entropy. In the process of image segmentation, this paper uses the method of the maximum image entropy based on fuzzy theory, and shows the segmentation method based on maximum fuzzy entropy is better by contrast the method of Otsu algorithm. In the part of identification and classification, the parameters of the defects were chosen by the area, the length, the breadth, the elongation and the compaction of the defects. This paper uses the BP neural network to classify the defects, and discusses the input and output data processing, the number of neurons and the choice of training methods in network construction. This paper simulate the identify network with the test samples, and gives the weight matrix and bias matrix of the trained network, and the output of some test samples. The simulation results show that the effect of defects classification is nearly 90%, and the research has good prospect.