论文部分内容阅读
研制了一种用于实时人脸识别的光电混合自动识别系统，包括特征提取、光照及几何归一化、编码、光学识别和判决几个部分。以计算机为控制中心，CCD摄像机实时获取待识别人脸图像后，由计算机做预处理工作，包括用灰度直方图均衡化做光照归一化，自动提取两个内眼角和嘴的中心三个参考点，利用仿射变换进行几何归一化并进行广义互补编码，以实现数学形态学击中击不中变换；编码图像送入基于非相干光相关器的光学处理器做光学运算；最后对光学识别结果进行判决，从图像库中找出一个或几个与待识别目标相似的人脸图像。给出了实验结果。
A hybrid optical automatic recognition system for real-time face recognition is developed, which includes feature extraction, illumination and geometric normalization, coding, optical recognition and decision. Taking computer as the control center, the CCD camera obtains the face image to be recognized in real time, and the computer preprocessing includes the normalization of the illumination with the equalization of the histogram of the grayscale, the automatic extraction of the three inner corners of the eyes and the center of the mouth Reference point, using affine transformation for geometric normalization and generalized complementary coding, in order to achieve a mathematical morphology hit hit can not transform; encoded image into the optical processor based on incoherent optical correlator to do optical computing; finally The optical recognition result is judged, and one or more face images similar to the target to be recognized are found out from the image library. The experimental results are given.