摘要
提出一种用Kohonen自组织神经网络进行雷达目标识别的新方法。这种方法可以对多个目标组成的多频模式空间向二维神经元阵列作拓扑有序映射.并提出一种输入输出神经元之间的连接权的微调方法——裂化映射算法,从而使自组织分类器的识别性能得到进一步提高。对两种飞机目标的性能进行了实验.结果表明,这种新的目标识别方法具有良好的应用前景。
A new technique to identify radar targets is presented which is based on a Kohonen self-organization neural network. It can map the pattern space of multiple-frequencies to a two-dimensional neuron array. A splitting algorithm to tune the weights between input and output nodes is also proposed. The performance identifying two airerafts by this neural classifier is studied and the results indicate that this new technique has bright prospects in the radar target identification.
出处
《北京理工大学学报》
EI
CAS
CSCD
1992年第3期67-72,共6页
Transactions of Beijing Institute of Technology
关键词
雷达目标
神经网络
目标识别
radar targets
self - organizaing control
multifrequency radar / neural networks
target identification
mapping
cracking algorithm
