摘要
本文首先在K-L变换中用相关矩阵作为特征判据进行特征压缩,然后用综合判别函数法制备空间综合匹配滤波器,可以有效地压缩特征图像数目.用这种滤波器实现了平移、旋转、尺度三重不变光学图像识别,且有较高的信噪比.
In this paper, we used correlative matrix as a feature criterion in K-L transformation for feature compression, and then the synthetic matched filter (SMF) was prepared by means of the synthetic discriminant function (SDF). This method can effectively redued the size of feature image. By this filter, the shift, scale and rotation invariant optical pattern recognition was realized with a high signal noise ratio of correlation output.
出处
《光学学报》
EI
CAS
CSCD
北大核心
1992年第1期52-56,共5页
Acta Optica Sinica
基金
国家自然科学基金资助的项目
关键词
信息处理
图像识别
匹配滤波器
information processing, optical pattern recognition, matched filter.computer-generated hologram, K-L transformation.
