基于模糊神经网络的海面目标战术意图识别

传统的基于遥感解译获取的海面目标信息分析战术意图的方法,需要大量的专家知识确定输入目标属性与输出意图间的网络节点关系,而模糊神经网络只需利用输入和输出训练网络,减少了专家知识的需求。针对常用的高木〖CD*2〗关野模糊模型不适用于战术意图识别要求的输出与输入非线性的问题,设计了基于神经网络集成的模糊系统模型,利用目标属性与对应的战术意图形成训练样本训练神经网络,分别获得输入条件的模糊隶属度以及面向不同意图的输出函数,据此识别海面目标战术意图。仿真实验结果表明,获得的目标战术意图的准确度高,与想定情况均相符。

Fuzzy neural network based tactical intention recognition for sea targets

Traditional methods of tactical intention analysis for sea targets, which are based on the information obtained from remote sensing interpretation, need plenty of expert knowledge to confirm the relationship of network nods between input target property and output intention. However, the fuzzy neural network only utilizes input and output samples in the network training process, reducing the requirement of expert knowledge. In tactical intention recognition for sea targets, the relationship between inputs and outputs is nonlinear so that the typical Takagi Sugeno model cannot handle this situation. Thus, a fuzzy system model based on integrated neural networks is established, in which target property and the intention are used to train neural networks to obtain the degree of fuzzy membership and output functions of different intentions. Using that model, the tactical intention of sea targets is then recognized. Experimental results present high accuracy of tactical intention recognition and are consistent with the situation.