基于卡方核的正则化线性判别行人再识别算法

针对行人再识别过程中存在获取的训练样本较少,真实样本分布不一定线性可分和算法识别率低的问题,提出基于卡方核的正则化线性判别分析行人再识别算法(KRLDA,kemel regularized linear discriminant analysis)。该算法首先利用核函数将样本从线性不可分的原始空间映射到线性可分的高维特征空间,然后在高维空间中构造描述数据之间邻近关系的散度矩阵,再利用正则化线性判别分析获得高维到低维空间的投影矩阵,使得数据在低维空间能够保持高维空间的可分性,从而提升行人再识别算法的识别率。在VIPeR、iLIDS、CAVIAR和3DPeS数据集上,实验结果表明所提出的算法具有较高识别率。

Chi square kernel regularized linear discriminant analysis for person reidentification

There are some problems in person reidentification, such as less training samples, no-linear relationship of samples, low recognition ratio. In order to solve these problems, the regularized linear discriminant analysis person reidentification algorithm (KRLDA) based on chi square kernel was proposed. Firstly, the algorithm mapped linear inseparable input data into a high dimensional linear separable feature space using kernel function to obtain scatter matrix that describes adjacent data relationship. Then, regularized linear discriminant analysis was applied to obtain low dimensional projection matrix for maintaining high dimensional separability characteristics to improve the recognition rate of the pedestrian re-recognition algorithm. Finally, experimental results on VIPeR, iLIDS, CAVIAR and 3DPeS datasets show that the proposed algorithm has a high recognition rate and the algorithm based on chi square kernel function has higher recognition rate than other kernel functions.