This work presents an application of Fractal Dimension (FD) and Segmentation techniques to Automated Visual Inspection. A system was developed for textile industries and aims at automatic failure detection. The textile industry environment presents particular characteristics in which usual morphological feature extraction cannot be used for visual quality control. Basically, the implemented system compares data from digital images; the characteristics of these data depend on the method selected for analysis. Two kinds of methods can be used: segmentation and fractal dimension. Segmentation techniques were adapted and improved to detect variations of the textile images (indication of defect). The fractal dimension is based on a new efficient algorithm that calculates the image fractal dimension values throughout all range that is between two and three (2 <= FD <= 3). These techniques were compared by their efficiency, accuracy and applicability.
Machine Vision; Automated Visual Inspection; Textile Industry
