改进Retinex算子融合特征搜索的图像识别算法研究  

Research on image recognition algorithm based on improved retinex and feature search

作　　者：李如平[1] 徐珍玉 

机构地区：[1]安徽工商职业学院电子信息系,合肥231131 [2]农业部农业物联网技术集成与应用重点实验室,合肥230088

出　　处：《齐齐哈尔大学学报（自然科学版）》2014年第6期14-18,共5页Journal of Qiqihar University(Natural Science Edition)

基　　金：安徽省对外科技合作计划项目(1403062028);安徽省科技攻关计划项目(13011031029;1201a0301008)

摘　　要：针对人脸图像识别系统易受光照干扰,在复杂背景下,易识别出伪目标,使其识别精度不高等难题,综合考虑高斯核函数的x,y方向,设计新的Retinex识别算子;并调用Open CV函数库中训练好的Hear分类器来进行模式匹配,形成了基于Open CV的人脸检测机制;再引入PCA特征搜索方法,搜索人脸信息;最终提出了人脸识别耦合特征搜索的安防系统解决方案。并测试该系统的性能,结果表明:与当前基于Retinex与PCA的识别算法相比,该算法具有更高的识别精度;且识别效率更高,能满足实时性要求。In order to solve these defects such as low recognition precision induced by pseudo-objective under the conditions of complex background which the face image recognition system prone to interference of light, the new retinex recognition operator was designed by considering the x, y direction of Gaussian kernel function;and pattern matching was done by trained hear classifier of OpenCV function library to inform the face detection mechanism. And then the face information was searched by introducing the PCA feature search method. Finally, the Security system solutions based on face recognition and feature search was proposed. Experimental results show that comparison with Retinex-based as well as PCA algorithms, this algorithm had higher recognition precision and recognition efficiency for meeting the requirement of real time.

关 键 词：人脸识别 特征搜索 图像 Retinex算子 PCA 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]