核函的LDA人脸图像识别.docVIP

基于核函的LDA人脸图像识别 摘要 对于人脸识别，大家似乎认为它离我们很远，其实人脸识别与我们的生活也是息息相关的。人脸识别是生物特征识别的一个主要研究方向，与其它生物特征识别技术相比较，人脸识别具有能动性，用户界面友好等许多特点。同时，在所有的生物特征方向，人脸特征也是最普遍和最比较容易获取的。因此，在模式识别和图像处理领域，人脸识别一直都是比较热门的研究课题之一。特征提取是模式识别学科研究的最基本问题之一。对于人脸识别而言，抽取有效的人脸特征是完成人脸识别任务的关键。从最初的基于几何的方法到基于统计等复杂特征的方法，人脸识别已经发展了很多算法。目前，基于统计特征的线性方法在人脸识别中以发展的比较成热，但是由于人脸识别涉及光照、表情、姿态等诸多因素，线性方法在实际也应用中表现得还远远不够。因此，将线性方法拓展到非线性领域以提高识别率是个有待解决的问题。基于核函数的特征提取方法是最近提出的一种很有效的非线性特征提取方法。本文就基于核的特征提取方法在人脸识别方面的应用进行了较深入的研究，所提出的算法在ORL人脸数据库上的试验取得了比较好的识别效率。 关键词：人脸识别；特征提取；核方法 Kernel Function-Based LDA Face Image Recognition Abstract For face recognition, everyone seems to think it is very far away from us, in fact, face recognition and our lives are also closely related. Face recognition is the biometric identification of a major research direction, compared with other biometric identification technology, face recognition book mobility, friendly user interface and many other features. Meanwhile, In the direction of all the biometric facial feature is the most common and most relatively easy to obtain. Therefore，In the field of pattern recognition and image processing, face recognition has always been a popular research topic. Feature extraction is one of the most fundamental problems of pattern recognition disciplines. For face recognition, the extraction of effective facial feature is the key to complete the recognition task. From the initial geometry-based approach to the methods based on statistical and other complex features, face recognition has been the development of many algorithms. At present, To develop linear methods based on statistical features in face recognition more into the hot face recognition involves illumination, facial expression, posture and many other factors, the linear method in practical applications, performance is still far from enough. Therefore, that the linear method is extended to the nonlinear field in order to improve the recognition rate is a problem to be solved. Based on multiple nuclear functions combination o