决策层时空信息融合的神经网络模型研究

融合目标识别可以获得比任意单传感器更加准确的目标识别结果。决策层融合由于其在信息处理方面具有很高的灵活性,成为信息融合研究的一个热点。针对传统融合算法环境适应性较差的缺点,提出了一种决策层时空信息序贯融合的神经网络模型,讨论了利用各传感器所处环境和专家知识等先验信息确定网络初始权值的方法,研究了网络权值的在线学习算法。仿真实验证明该网络模型的有效性。

Research on a new network model for temporal-spatial information fusion at decision level

Information fusion for target recognition can generate more accurate classification result than each of the constituent sensors.Because of the high information processing flexibility,fusion at decision level has become a research focus on information fusion.Aiming at the shortage of the common fusion scheme which cannot adapt itself to environment change,a new neural network model for temporal-spatial information fusion at decision level is put forward.The expert knowledge and environmental information are sufficiently used to initialize the network,and an on-line learning algorithm for the network's connected weights is given.Simulation result shows the efficiency of the new network's model.