Automated defect detection in textured surfaces using optimal elliptical Gabor filters

In this paper, a new approach is presented that addresses defect detection in textured surface by using an optimized elliptical Gabor filter (EGF). This newly proposed EGF can be tuned by a genetic algorithm (GA) to have either a form of traditional Gabor filter or a form of shifted ring Gabor filter, with the aim to match with the texture features of a defect-free template acquired in prior during the training process. In the inspection process, each sample under inspection is convoluted with the selected EGF, followed by a gray level thresholding process to generate a binary segmented result. Since only a single filter is utilized during the inspection process, the computational time required for each defective sample is quite limited when compared with those filter bank based schemes. The performance of the proposed method has been extensively evaluated by a variety of samples with different defect type, shape, size and texture background. Experimental results demonstrate the effectiveness of the proposed method on defect detection in textured surfaces.