Image processing based fault dedection system in manufacturing automation

At the last decade, traditional industrial systems are converted to full automated industrial systems rapidly. Hence, demand of quality and efficiency has been increasing and quality control has been getting importance at the production. Quality control has been performed by human observation with low efficiency and slowly. Instead of human inspection, digital system based quality control presents high efficiency and high speed at the full automated industrial systems.In this study, we have developed scratch detection and classification algorithms based on morphologic operators, adaptive thresholding, and pursuing algorithm for video sequences. This study achieves fault detection rapidly without hampering the manufacturing process. The proposed scratch detection system based on image processing techniques is implemented at four steps. First stage is recording product image by a suitable camera. In the next stage, random positioned product in the scene is registered to correct coordinates by devised scanning algorithm. In order to operate rapidly in the next stages, image backgrounds are removed by purification algorithm and RGB images are converted to gray level images. In stage three, Canny algorithm have been applied to reveal the scratch on the surface. In this process, the critical challenging problem is the determining of the threshold. To overcome this problem, we have utilized Otsu adaptive thresholding method which offers better solution for scratch detection. Morphological operators are then incorporated to restore the break points stemmed by thresolding. According to size and numbers of dedected scratchs, the product classified into two categories; faulty products or acceptable product.