D-S理论与神经网络相结合的信息融合模型研究

针对大噪声环境下信息融合效果差的问题,提出了一种基于D-S证据理论与神经网络技术的信息融合方法,该方法综合了证据理论在处理不确定信息方面的优点和神经网络在数值逼近上的长处,一方面利用神经网络和冲突证据处理算法获取基本概率赋值,另一方面通过证据理论使神经网络的结构变得透明.初步仿真结果表明,该方法有效地解决了不确定性信息的误识别问题.

Research of Information Fusion Model Combining D-S Evidence Theory and Neural Nerwork

Under the circumstance of great noise in information fusion system, the fusion effect is poor. In view of this problem, thus a method of information processing is proposed based on the combination of D-S evidence theory and neural network. Evidence theory has the advantage of processing uncertain information while the neural network is ascendant in numerical approximation. In the method, by integrating both advantages, the basic probability evaluation can be obtained using neural network and the processing arithmetic of conflict evidence. Moreover, the architecture of neural network becomes apparent according to evidence theory. Simulation shows that the recognition rate of uncertain information is improved.