Neps detection for wool fabrics by using image processing techniques

Image processing methods, started to be used for military targets in 1960s, have found usage opportunities in many areas thanks to advanced and fast computer technology. Textile is one of these areas. The process, made on the thesis includes efforts to identify neps on wool fabrics by means of on-line image processing using line scan cameras. Neps and foreign substance on fabric surface affect negatively quality of fabric. In the thesis image processing methods were analyzed and the most suitable method was determined for on line image processing of fabrics. Most of the image processing studies are related with the inspection and classification of fabric faults, however, this study aims on-line image processing to determine neps and their locations to be used for neps cleaning process by a robotic machine. As a consequence, the most suitable image process method was decided by systematically eliminating the others for neps type fabric faults. The selected method was practiced on a computer controlled neps cleaning robot for on line image processing and neps cleaning purpose. The process procedure starts with taking image of the fabric (1 meter length and 1.6 meters width) by means of two line scan cameras. The image then processed to determine faults (neps) and their locations. The location information is used as input data for robotic machine to achieve the coordinates of neps. Then, neps cleaning robot moves to the related coordinates and removes them in a systematic order. The scope of the thesis is limited with the analyzing the image processing methods and applications of the most suitable (fast and effective) method for the image processing.