计算机工程与应用 ›› 2012, Vol. 48 ›› Issue (1): 205-207.

• 图形、图像、模式识别 • 上一篇    下一篇

复杂背景下的人脸定位识别方法

仲 澄1,冯 涛2   

  1. 1.上海中新科技管理学院,上海 200232
    2.上海第二工业大学,上海 200041
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2012-01-01 发布日期:2012-01-01

New algorithm for face location and recognition

ZHONG Cheng1, FENG Tao2   

  1. 1.Science and Technology Management College, Shanghai 200232, China
    2.Shanghai Second Polytechnic University, Shanghai 200041, China
  • Received:1900-01-01 Revised:1900-01-01 Online:2012-01-01 Published:2012-01-01

摘要: 现有的人脸识别算法多在标准库上进行,缺少对复杂背景下人脸识别问题的研究。提出一种快速的人脸定位识别方法,旨在解决复杂背景中人脸的定位和识别问题。在定位方面,提出一种新的自适应肤色分割的人脸定位算法,充分考虑类肤色背景对定位算法的影响,使该算法在户外环境下的人脸定位精度较传统方法有了一定的提高;识别方面,采用局部SVD方法提取人脸图像特征值,以PCA算法加以识别,新算法改进了传统PCA训练速度慢、内存占用大的缺陷。通过对ORL人脸库以及自制人脸库的实验分析,结果表明该方法不仅能解决复杂背景中人脸定位识别问题,并且高效、快速、有较好的实用性。

关键词: 人脸定位, 人脸识别, 自适应, 主成分分析(PCA), 局部奇异值分解(SVD)

Abstract: Most existing face recognition algorithms are based on standard face database. There is only a little face recognition research about complex background. This paper proposes a new face location and recognition algorithm to solve the problem. Face location is based on adaptive skin color segmentation. By taking similar skin color into consideration it can achieve face location of pictures under complex background. Face recognition uses area SVD to advance the PCA method, improves the recognition rate and speed. Through the test of self-face database and ORL database, the efficiency and robustness of the algorithm have been verified. What is more important, the new algorithm achieves face recognition of pictures under complex background.

Key words: face location, face recognition, adaptive, Principal Component Analysis(PCA), area Singular Value Decomposition(SVD)