The Application And Research Of Fractal Dimension In Leather Classification

Our country is the leather industrialized nation, but the classification of leather is still often relying on visual identification of skilled workers. It is difficult to ensure consistency and stability of the product. With the development of electronics and computer technology, we can adopt the picture processing technology to transform the leather visual information into the numerical information, and using pattern recognition technology allows automated classification of leather possible.Texture analysis techniques are used in this dissertation, and fractal dimension is used as the main feature of texture of leather image, combined with the threshold method and neural net work, to study the classification of leather image. The main contributions of this work are as follows:This paper researches three methods of fractal dimension calculation, which called box counting method, differential box counting method, and blanket covering method. Formed the solution program, which is about how to calculate the fractal dimension of leather images.In order to use these methods, the paper detailed analysis the characteristics of the leather image, presents a series of pre-processing to leather image, including monochrome image and gray image. Histogram equalization is used to enhance the leather images, highlight the difference between the images, and enhance the feature of the images.A classifier of leather images based on fractal dimension and threshold method is designed. This classifier is used to distinguish the leather image, combined with three kinds of fractal dimension. Do some experiments about the classification of leather image, and determining the threshold. The results of experiments are discussed and analyzed in detail.The classification method of leather images, which based on curve fitting and the fractal dimension of image block, is used to distinguish sheep skin image and goatskin image. Do some experiments about the classification of leather image. The results of experiments are discussed and analyzed in detail.In order to improve classification accuracy, this paper using the method of the classification of leather image based on BP neural network and fractal dimension feature vectors, to do some classification experiments of leather image. The results of experiments are discussed and analyzed in detail.