Detecting and Classifying Defects in Textile Fabrics with Gabor Filters and Neural Network

Abstract

Given the importance of the industry and because the quality of the products reflect the promoted and progress and the advancement of the economy of any country , including the textile industry one of the most important industries that require quality. As the methods of automation was necessary to use computer vision and image processing to increase the speed and efficiency of this process. Aims of the proposed work to integrate the image processing methods and intelligence techniques as well as statistical approaches. Where analysis techniques multiple scales multi-scale and multi-directional multi-orientation as a filter Gabor are used. And this filter has proven its efficiency in edge detection and give the best features by which they are distinguished types of defects which you may get during the spinning and weaving . Hence the formulation of how they can address the installation of textile and defective areas of the cloth and to identify any kind of flaws in them.To raise the level of this process when checking woven-fabrics and identify defects. The proposed work includes two phases, the first phase is to detects detection in fabric images and the second stage is the stage of classification defects. At the separation phase image is converted into frequency space by conversion sinus intermittent (DCT). Then features are extracted and inserted into the Backpropagation neural network where the separation process is done. Either at the stage of classification are images are converted into frequency domain by Gabor transmition .And then draw features images are inserted into the Backpropagation neural network to classify fabric defects in those images .To verify the efficiency of the techniques used, live images were collected as a database of fabrics from the textile laboratory in Mosul as well as the local market. The fabrics were carefully chosen with fabrics of different types and colors and fourteen different types of Fabric defects. System was used (Matlab 2013). Explained the proposed work discrimination ratio ( 97.5%) compared to the results of the other works in the same field approach.