Modification algorithm of DS evidence theory based on the evolution function of focal elements' energy
Chen Ying-chun, Sun Young Yu, Li Ou · 2017
Dempster-Shafer(DS) evidence theory can usually distinguish the uncertain information and the unknown information in a weaker constraint condition than the probability theory and Bayes' theorem. It is an efficient tool in information fusion, target recognition and decision-making under uncertain environments. But, the high conflict between evidences can generate fusion results contrary to practice in DS evidence theory. It is unavoidable that high conflict evidences are caused by noise or interference in practice, especially for the sensor data. In order to decrease the influence of the high conflict and noise or interference on the fusion results, in this paper, a new modification algorithm is proposed to improve the performance of DS evidence theory by modifying the evidence sources and the combination rule together. Firstly, in order to make full use of not only the basic probability assignments (BPAs) but the cardinality of focal elements, in the modification algorithm, a novel concept called the evolution function is defined in the form of the prior information ratio of focal elements' energy. Secondly, with a threshold of the average conflict coefficient between evidences, the evidence sources are modified by using the evolution function. Finally, based on the conflict information and the prior information, a modification method of the combination rule is presented by using the evolution function and the conflict coefficient in the modification algorithm. The experiment results prove the accuracy and the effectiveness of the proposed modification algorithm.