基于计算机视觉的人脸识别算法的研究  

Design of Security System Based on Computer Vision

作　　者：肖彬 李泽滔[1] 杨昱翔 XIAO Bin;LI Ze-tao;YANG Yu-xiang(The Electrical Engineering College of Guizhou University, Guiyang, Guizhou 550025)

机构地区：[1]贵州大学电气工程学院,贵州贵阳550025

出　　处：《新型工业化》2018年第12期61-66,共6页The Journal of New Industrialization

摘　　要：非限制人脸识别技术是身份识别领域研究的核心内容,其具有隐蔽性高、侵犯性弱和实用性强等特点。但是日常需求的非受控环境,如光照、姿态、遮挡等因素,会导致人脸的识别性能急剧下降。针对目前非受控环境下人脸识别的问题,本文提出一种有效可行的思路:利用LBP(Local Binary Patterns)算法的良好对光照的鲁棒性,结合DBN (Deep Belief Network)对表情、遮挡等的良好鲁棒性,增强人脸识别的识别精度。同时本文通过在相同人脸库的条件下,与其他的各类算法进行对比,可以看到本文提出的算法明显优于其他的算法。Unrestricted face recognition technology is the core of the research in the field of identity recognition. However, the uncontrolled environment of daily needs, such as lighting, posture, shielding and other factors, will lead to a sharp decline in face recognition performance.Aiming at the problem of face recognition in the uncontrolled environment at present, this paper proposes an effective and feasible idea: use the good illumination robustness of LBP (Local Binary Patterns) algorithm and combine the good robustness of DBN (Deep Belief Network) on expressions and occlusion to enhance the recognition accuracy of face recognition.At the same time, the algorithm proposed in this paper is obviously superior to other algorithms by comparing with other algorithms in the same face database.

关 键 词：计算机视觉 人脸识别 LBP算法 DBN 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]