The Research Of Leather Defect Detection Using Texture Analysis Techniques

With the development of people's living condition, the leather productions increase rapidly, people like them very much. Consequently, people pay more attention to the quality of leather. On the other hand, the speed and the precision are not satisfied to social progress through unaided eye sight, the result is vary because of individual subjectivity of decider, there is no uniform standard, and the speed and the precision will be decreased through repeated work. With the broad application of computer image technology, the inspect level of leather quality increased fleetly. By the means, the pressure of workers can be reduced and the labor productivity can be raised. The speed and precision will reach a high level. Through the computer image technology, we can inspect the leather quality from the leather surface defects, the leather's flatness and the density of leather, and so on. The most important aspect is the surface defects. The article main inspect the defects of leather surface and evaluate defects use the computer image technology. Therefore, the work of the assay is as follows:For improving the precision of leather segmentation and defects detection, the leather need to be pretreated. The article analyzes the characteristic of leather which have defects. The pretreat course includes reinforce the contrast of image and denoise to image. There are many methods to reinforce image, it needs research best way for leather reinforcement. The texture and noise of leather are high frequency information; a best mean will be investigated for the purpose of removing the noise and keep the texture information. After comparison, some best ways are used to leather pretreatments, the means of gray linear transform is used to reinforce the contrast of image and the means of wave packet transform is used to remove the noise. And the step and best way to pretreat the leather are suggested.In the phase of leather segmentation and the surface defects detection, the methods of segmentation for gray image and texture image are compared. It analyzes that the defects are not obvious, the gray of defection regions and non-defection regions are similar. The leather is natural texture image, the defection regions and non-defection regions have biggish difference of texture, so the texture method is used to detect defects. The aid is to select a best way to separate the defects from good regions, or several methods are combined to reach a perfect effect. Some texture characters are extracted from leather surface use gray level co-occurrence matrix and the image is segmented use the optimized Fuzzy C-means. The optimized Fuzzy C-means add the spatial information; as a result the precision of segmentation is improved. The image needs to be treated use morphological approach after it is segmented. As a result, the defective regions are separated from no defective regions successfully.During the course of evaluating the leather surface defects, many facts are considered. It includes type, number, location and area of the defects. According to the distribution of defects gray level, a method to estimate the type of each defect and count the number of defects are suggested, it provides scientific proof to evaluate the quality of leather.Finally, this article work is summarized and the future work is prospected.