A Research On Decision Information Fusion And Military Application Based On Evidence Theory

With the development of information technology and the improvement of the complexity of modern war, various factors push the information fusion technology ahead rapidly. Evidence Theory, as a deduction of uncertainty in the field of information fusion, is being wildly applied many fields, especially in target recognition fusion. Though it has developed for more than40years, it still has many drawbacks. This thesis studies the measure and synthesis of conflictive evidence and the application of target recognition in the sky.The thesis studies the trust interval and uncertainty, and analyzes the meaning of some common trust intervals; it also studies the combination rule of D-S Evidence Theory. The basic principle is analyzed and popularized to the combination of multiple evidences. Meanwhile, two problems exist. Firstly, when dealing with the evidences of high conflict, paradox may appear. Secondly, during the process of combining, the trust value in multi-point set cannot be passed on reasonably to its subset, which means the trust value cannot be allocated reasonably. Two examples are listed to analyze and embody the two questions.The thesis studies the method of measure and combination of conflictive evidences. Due to the fact that the classical conflict coefficient only includes the non-inclusive of evidences, the thesis introduces the potential conflicts, analyzes the conflicts of compatibility focal elements, improves the classical conflict coefficient, and then brings forward a new method of conflict measure of conflictive evidences by combining with the Jousselme distance function. After comprehensive consideration of the sufficiency and importance of evidences, it improves the combination method of conflictive evidence. On a new measure, by forming a new discount factor, considering the sensors of different weighing will lead to different importance of evidence, introducing rough set getting sensors’ weighing, and by using the two coefficients to revise evidence sources, it puts forward a method of combination based on modified evidence model.It was studied the target recognition based on the improved D-S Evidence Theory. The integration technology of target recognition is studied in three levels. The working principle of conventional target recognition model is analyzed. According to the previous research of D-S Evidence Theory, combining the present equipped establishment with performance, and improving the conventional target recognition model, the thesis brings forward a two distributed fusion model of target recognition based on the improved D-S Evidence Theory and studies the main process of target recognition based on the improved D-S algorithm. The method and model is proved to be practical and reasonable by examples.