Fusion Recognition Model Of Marine Targets Based On Improved D-S Evidence Discount Theory

The emphasis of this dissertation is put on the construction of recognition model based on the sensors' output recognition vector. The main method is improved D-S discount evidence theory. When solving the problem of fusing evidences with reliability, two defects of D-S Evidence discount theory is exposed. One is that results of original algrithm vary according to the content of evidences. The other is that original algorithm can't adapt to the changing environment and can't express men's will when fusing evidence with reliability. An improved algorithm is presented. According to the single relationship between the evidence reliability and its affected factors, a fuzzy neural network that has a special structure is built. FNN is used to simulate the experts' experience to get the evidence reliability. The study algorithm is also presented. In the output part of the model, the risk function rale is introduced to make a decision in some certain conditions. At last, maritime target recognition is given in a small-scaled simulation to check the model. The results proof that the model can make a proper decision when fusing evidences with reliability, which is useful in the theory research and real recognition fusion. The future research aspect is to mcdify the model details and improve the fusion level of the model.