Data Fusion And Data Mining Theory Applied Research In Target Recognition

Underwater target recognition, which possesses critical application value to state economy and defense, is an important technique of modern sonar and acoustic counterwork. In this dissertation, the concepts and principles of data fusion, data mining and target recognition have been comprehensively discussed. Aiming at the detailed problem of underwater target recognition, we give thoroughly studies on techniques of data fusion, data mining and their integration.Firstly, the constitution and common techniques of underwater target recognition system are analyzed. The basic theory of data fusion and data mining is introduced. And the effect of the two theories in target recognition is discussed.Secondly, Dempster-Shafer evidence reasoning, an important research technique of data fusion is thoroughly discussed including its basic concept and synthesizing principle. A D-S fusion model of target recognition is founded for the detailed application of underwater target recognition. The forming mechanism of the radiated noises of ships, submarines and torpedoes and the basic techniques of feature extraction are discussed. In a practical simulating experiment, three feasible project are developed based on three different methods of achieving basic probability assignment. The comparison and summarization of experimental results are listed.Thirdly, Rough Set theory, a tool of data mining is imported. A series of processing from establishment of information expression system, dispersion and reduction of decision table to creation of decision rules are elaborated. Two aspects of application of Rough Set theory in underwater target classification and recognition are put forward. One is achieving superior decision rule to improve ratio of recognition by using related Rough Set reduction algorithm. The other is using the concept of property importance measurement to guide the key-point of target recognition feature extraction. Finally, the similarity and difference between data fusion and data mining are further discussed. As two techniques analyzing uncertain information and extracting useful knowledge, data fusion and data mining have different purpose, principles and techniques, but they can complement each other in function. In this dissertation, the integration of data fusion and data mining is discussed. The combination of Rough Set theory and fuzzy logical theory is realized. The relationship of Rough Set theory and D-S evidence reasoning is analyzed, and they can combine with each other by establishing a specific D-S fusion model considering weight of the importance of evidence.