DETECTION OF LOUSINESS IN SILK FABRIC USING DIGITAL IMAGE PROCESSING

An inspection method based on binary image processing technique has been developed in an effort to obtain lousiness profile of silk fabric. The image data are used in the mathematical morphology and the back propagation neural network (BPNN) for evaluation. Maximum length, maximum width and grey level fabric defects are considered as input units in the input layer of BPNN. The grey level values corresponding to the image pixels are used to perform recognition of three types of defects namely, lousiness (high), lousiness (moderate) and lousiness (low) successfully with frequency of each defect per unit area. The average recognition rate is found to be 98.56 % under 3-3-3 BPNN topology. The results prove that the inspection method developed is very effective not only in identifying and recognizing the defects of very small size like lousiness but also act as a tool in generating almost zero-defect finer fabric.