用多层前向网络设计雷达目标分类器

Designing Radar Target Classifier by Means of Multilayer Feedforward Neural Network

概略地讨论了基于高分辨力技术的雷达目标识别问题。针对雷达目标分类器的技术需要，着重研究了多层前向网络的分类性能，提出一种网络结构自适应策略。并用其设计了分类器，在转台成像的基础上，对飞机目标的类型识别问题进行了仿真研究。研究结果表明，小规模多层前向网络对基于距离像的雷达数据样本具有较好的推广识别能力，识别率在９０％上下。

The technology of radar target identification (RTID) based on high resolution radar is discussed briefly. The classification performance of multilayer feedforward neural network (MFNN) is investigated, and a structural adaptative strategy is presented in this paper. Then several simulations on RTID have been taken by using MFNN to design a classifier. The results demonstrate that a small scale MFNN has better generation performance after being trained by range profiles, and the right recognition rate is up to 90% or so. [WT5”HZ〗