齐齐哈尔大学学报（自然科学版）
齊齊哈爾大學學報（自然科學版）
제제합이대학학보（자연과학판）
JOURNAL OF QIQIHAR UNIVERSITY(NATURAL SCIENCE EDITION)
2014年
6期
14-18
,共5页
人脸识别%特征搜索%图像%Retinex算子%PCA
人臉識彆%特徵搜索%圖像%Retinex算子%PCA
인검식별%특정수색%도상%Retinex산자%PCA
face recognition%feature search%image%Retinex operator%PCA
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OpenCV函数库中训练好的Hear分类器来进行模式匹配，形成了基于OpenCV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。
針對人臉圖像識彆繫統易受光照榦擾，在複雜揹景下，易識彆齣偽目標，使其識彆精度不高等難題，綜閤攷慮高斯覈函數的x，y方嚮，設計新的Retinex識彆算子；併調用OpenCV函數庫中訓練好的Hear分類器來進行模式匹配，形成瞭基于OpenCV的人臉檢測機製；再引入PCA特徵搜索方法，搜索人臉信息；最終提齣瞭人臉識彆耦閤特徵搜索的安防繫統解決方案。併測試該繫統的性能，結果錶明：與噹前基于Retinex與PCA的識彆算法相比，該算法具有更高的識彆精度；且識彆效率更高，能滿足實時性要求。
침대인검도상식별계통역수광조간우，재복잡배경하，역식별출위목표，사기식별정도불고등난제，종합고필고사핵함수적x，y방향，설계신적Retinex식별산자；병조용OpenCV함수고중훈련호적Hear분류기래진행모식필배，형성료기우OpenCV적인검검측궤제；재인입PCA특정수색방법，수색인검신식；최종제출료인검식별우합특정수색적안방계통해결방안。병측시해계통적성능，결과표명：여당전기우Retinex여PCA적식별산법상비，해산법구유경고적식별정도；차식별효솔경고，능만족실시성요구。
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.
