By analyzing the metal surface microstructure, people can know more about the capability and usage of the metal. Metallographic images are images of metal microstructures. By image analysis techniques, metal lographic image analysis can be developed from human observing and qualitative analysis to automatic and quantitative analysis.In accordance with the features of metallographic images, a useful metal lographic image analysis system has been constituted, combined with computer technique, image analysis technique and pattern recognition technique. In this thesis, we have carried out some researches of texture feature extraction algorithm and classifier. There are many algorithms of texture feature extraction. It is good to combine the unified probability matrix algorithm and wavelet transform algorithm to extract the texture features. We also discuss the Mahalanobis distance classifier, and develop it with confidence level and voting system. It can get higher precision. BP neural network is one of the most widely applied artificial neural networks. So it is worth studying how tointroduce BP neural network into metal lograghic image analysis system. It is difficult to analyzing a metal lograghic image in a conventional way, if the background and the objects can not be segmented or the objects are hardly isolate. If applying, however, the texture analysis technique in the system, this problem can be solved well.
|