Research And Appication Of Image Recognition Based On Kernel-Based Learning Algorithms

Automatic fiber recognition deals with many research fields such as image processing,artificial intelligence.Because the manual fiber recognition is time-consuming and its accuracy is largely depended on operators,the computer aided fiber recognition has received much attention.However,the algorithms for automatic fiber recognition are complicated,and there are no successful research results.This paper is a part of the research sponsored by the Foundation of National Excellent Doctoral Dissertation of China and the Scientific Research Foundation for the Returned Overseas Chinese Scholars,State Education Ministry. The research is also sponsored by the Shanghai Entry-Exit Inspection and Quarantine Bureau of P R China.The research mainly deals with the algorithms for automatic recognition of the nature cellulose fiber and the classification of shaped fiber.The paper pays attention to recognizing the shaped fibers by computers.As the limits of the production process and shooting conditions of the fiber samples,the quality of the fibers' cross-section images are too poor to be recognized in conventional ways.Kernel learning algorithms have been presented in the 1990s.Nowadays they are rapidly developed. They belong to a pop-researching field internationally and arouse extensive attention and upstairs recognition in science and technology field.The representation kernel-based learning algorithms-Support Vector Machines(SVMs) and Kernel Principal Component Analysis (KPCA) have been reported for various fields,for instance in the pattern and object recognition,time-series prediction,fault detection.But these algorithms have been rarely used in the field of textile.The paper first attempts to introduce KPCA and SVMs to the field of the textiles' automatically recognition.The algorithm of recognizing shaped fibers based of the KPCA and SVMs is presented by the paper.It has been implemented in the shaped fibers recognition system.The algorithm extracts the features from a gray-scale image of a single shaped fiber by KPCA.When the features are the input of SVMs,the classification result will be outputted by SVMs.KPCA extracting high-order statistic information from origin gray-scale image is interfered by the quality of the image.SVMs provide a stable classifier with generalization.The paper discusses modules and functions of the recognition system.With the recognition system, comparison experiments are carried out on the kernel functions.The result is that polynomial kernel function is better than BRF and sigmoid function.Meanwhile,the effect of the order of the feature space on the performance of the recognition system is analyzed through the experiments.The paper first introduces the kernel-based learning algorithms KPCA and SVMs to the fields of the textiles' automatically recognition. The algorithm of the paper for the shaped fibers' microscopic images recognition is proved to perform well.The paper provides a new way of the research of the shaped fibers' automatically recognitions technology.