基于LBP-HSV模型及改进SIFT算法的行人再识别算法

Person Re-identification Algorithm Based on LBP and HSV Model and Improved SIFT Algorithm

摘要:

针对SIFT算法计算量大,复杂背景下匹配准确率低的问题,论文提出了一种结合LBP-HSV模型与改进SIFT算法的目标识别算法。首先利用LBP直方图和HSV模型共同筛选出目标相似区域；然后利用SIFT算法检测目标与相似区域的特征点,并使用改进的HOG特征描述特征向量；最后采用最近邻加权欧式距离的匹配策略,找出匹配点对。基于多组行人图片的目标识别结果表明,文中算法具有较强的鲁棒性,识别准确率较高,且相较于SIFT算法,匹配速率大大提高。

Abstract:

Aiming at solving the problem that SIFT (scale invariant feature transform) algorithm’s computation complexity is high and its matching accuracy is low in a complex background, an image matching algorithm by combining LBP-HSV model and improved SIFT algorithm is proposed. It first utilizes LBP histogram and HSV model to screen the identified similar region. Then it uses SIFT algorithm to detect the feature points of the target and alternative region, and take advantages of improved HOG feature to describe feature vectors. Finally, it finds matching points by using k-nearest-neighbor algorithm and weighted Euclidean distance. The results of experiments carried out on multiple pedestrian pictures show that the proposed algorithm has good robustness and high accuracy, and compared with SIFT algorithm, the matching speed is greatly improved.