用于人脸识别的非相干光形态学处理器  

Object Recognition by the Use of an Incoherent Optical Morphological Processor

作　　者：成罡[1] 金国藩[1] 邬敏贤[1] 袁石夫[1] 宋修宇[1] 

机构地区：[1]清华大学精密仪器系,北京100084

出　　处：《高技术通讯》1997年第4期41-44,共4页Chinese High Technology Letters

基　　金：863计划;国家自然科学基金

摘　　要：设计了一个用于人脸图像光学识别处理的非相干光形态学处理器。系统中采用互补编码击中击不中变换以改善传统光学相关器的性能。基于该方法，构建了一个紧凑性的人脸光学识别非相干光系统，图象库中存有96幅人脸图象，可以达到每秒钟10幅的识别速度，识别正确率大于90％，对于8°以内的旋转畸变、30％以内的Gauss噪声干扰以及50％以内的图像缺损具有相当好的容错能力。A scheme for the optical implementation of human- face recognition with an incoherent optical morphological processor is presented. The system uses complementary -- encoding hit- miss transform method to improve the performance of the standard correlator. According to this method, a compact system for human-face optical recognition is built up. A face library is built in which 96 photographs are stored and the recognition speed of the system is 10 pictures per second. Grey level map pictures have also been recognized in the experiment.

关 键 词：人脸 光学识别 互计编码 数学形态学 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]