基于algtp的人脸识别研究-计算机软件与理论专业论文.docxVIP

Abs Abstract Shanghai Normal University Master of Philosophy 万方数据 万方数据 Abstract Face recognition technology has been widely used in public safety and daily life, etc, and it is becoming one of the hottest topics in the field of pattern recognition and computer vision in recent decades. But in practical applications, the face images are acquired in the non-constrained condition, and the number of the training samples is limited. Under this circumstance, the performance of the face recognition system will be seriously affected. The research on the most effective and discriminative face feature extraction algorithm under various non-constrained condition is the key to the development of robust and efficient automatic face recognition system. Local Binary Pattern (LBP) is an efficient local feature description operator. In recent years, it has received more and more attention and has been successfully applied in face recognition. But LBP itself is still not perfect and there still exists several problems: Due to the totally dependence on the center pixel of the local area for the value of threshold, LBP is sensitive to noise. Furthermore, the uniform block partition before LBP feature extraction will cause the deviation. And the performance is not robust enough under complex light illumination conditions. This paper focuses on these issues. The main work of this paper is as follows: Aiming at the problem for the value of the threshold for LBP Operator which totally depends on the center pixel of the local area, ALGTP (adaptive local Gabor ternary pattern) method for face recognition is presented in this paper. By calculating the mean and variance of the local area, the method automatically generates the threshold of local area, which no longer completely depends on the central pixel of the local region. To solve the problem of the deviation caused by the uniform block before the image feature extraction, a more robust algorithm based on SIFT algorithm is proposed, which