摘要
研究提高人脸识别率问题,因人脸图像易受光照条件、人脸丰富的表情变化以及周围复杂环境干扰等因素的负面影响,导致其识别准确度很低,影响其识别效果。鉴于此,提出了改进型PCA和LDA融合算法人脸图像识别方法,首先通过在改进PCA算法中结合基于标准差和局部均值的图像增强处理,使其可以有效调节光照不均匀对人脸识别所造成的负面影响,进而拓展了PCA算法的应用条件范围,然后将改进的PCA算法与LDA算法相结合,运用改进的PCA算法对训练图像降维,最后再对降维以后的特征采用LDA算法,训练出一个最具判别力的分类器,实验证明本文提出的方法对光照不均匀、表情变化的人脸具有一定的鲁棒性,具有很好的人脸识别性能,提高了其识别率,优于一般的PCA算法。
To improve the face recognition rate, in this paper, we presented a face image recognition method by integrating an improved PCA and a LDA algorithm. Firstly, to improved PCA algorithm, we introduced the image enhancement processing based on standard deviation and partial means, in order to effectively regulate the affects of uneven illumination for face recognition. And then, we extended the application condition limits of PCA algorithm, and integrated the improved PCA and the LDA algorithm. We made dimension reduction by using the improved PCA algo-rithm for training image, and applied the LDA algorithm for the feature after reducing dimension, in order to train the most discriminative classifier. Experiment results show that the proposed method has a better robustness for uneven il- lumination and the changes of face expressions, and face recognition performance has improved significantly compared with the general PCA algorithm.
出处
《计算机仿真》
CSCD
北大核心
2013年第1期415-418,426,共5页
Computer Simulation
基金
新疆大学校院联合资助项目(XY110133)
国家自然科学基金(60865001)
关键词
人脸
全局特征
鲁棒性
Face
Global features
Robustness