The Non-destructive Testing Defect Pattern Recognition Based on Information Fusion

This paper attempts to adopt multi-sensors system based on Dempster-Shafter evidential reasoning approach and the corresponding multi-source information fusion technology, to overcome the shortcomings that it can't understand unknown environment and to study requirement accurately and comprehensively. Firstly, this paper constructed basic probability distribution function, the basic probability distribution functions of different evidence bodies are merged into a total of basic probability distribution functions. Secondly, the classification decision strategy was presented. The strategy includes four rules which are used to solve the analysis and decision of basic probability distribution function. Thirdly, a multi-source information fusion system which has data pre-processing, information fusion, classification and recognition functions was presented. Lastly, the experimental research on defect was done, the result obtained from the experiment shows that the classification method with information fusion is better than single invariant recognition method.