摘要
本文针对雷达地图匹配制导中的共性特征提取和雷达与地图匹配这一特定问题，应用小波变换多尺度分析提取共性特征．提出基于人工神经元网络的雷达地图匹配方法，并且给出相应的试验结果．因为Ｈｏｐｆｉｅｌｄ 模型可以由集成电路实现．本文算法可以实时地完成．试验表明采用本匹配方法能够解决雷达地图匹配问题，获得比传统匹配方法。
In this paper relaxation matching is investigated as a technique for matching map features to features in the SAR images.A method that applies wavelet transform to extract a number of feature points as a basis for matching and makes the Hopfield neural network perform the mutual feature point relaxation matching process is proposed.In such a way,the image matching by the relaxation process can be realized in real time since the Hopfield model net can be implemented by VLSI.The experimental result has showed that our approach is more efficient and has higher matching probability than usual approaches based on gray level and edges,such as MAD (Mean Absolute Difference) algorithm.
出处
《电子学报》
EI
CAS
CSCD
北大核心
1999年第10期58-61,共4页
Acta Electronica Sinica
基金
国防科技预研基金
航天基础性研究基金
中国科学院自动化所模式识别模式识别国家重点实验室基金
关键词
雷达图像
多尺度分析
共性特征点
松弛匹配
radar image
analysis in multiscale
mutual feature points
neural network
relaxation matching
