基于典型相关分析特征融合的行人再识别方法

A person re-identification method based on feature fusion of canonical correlation analysis

传统的行人再识别方法通常使用手工设计的视觉特征来描述行人图像。然而,仅用单个类型的视觉特征很难全面表征行人图像信息,导致识别性能达不到满意效果。提出一种利用典型相关分析对不同类型特征融合的行人再识别方法。该方法首先分别对两种不同类型的行人视觉特征进行典型相关分析,以获得两组不同特征之间的最大相关子空间。然后,分别使用连接融合和相加融合两种策略对两种变换后的特征进行融合,使用融合后的特征用于行人再识别。实验结果表明,相比单个类型的行人图像特征和简单的多特征连接方法,提出的特征融合方法在保持特征维度较低的同时,能获得更好的识别率。

Traditional person re-identification(Re-ID)methods usually use handcrafted features to describe pedestrian images.However,it is difficult to fully characterize the information of pedestrian images by single kind of features,resulting in unsatisfactory recognition performance.A person re-identification method that uses canonical correlation analysis(CCA)to fuse different kinds of features is proposed.First,the proposed method obtains the maximal correlative subspaces of two different types of visual features by using canonical correlation analysis.Then two fusion schemes,i.e.,the connecting fusion and the summing fusion,are applied to combine the two kinds of transformed features for person re-identification.Experimental results indicate that the proposed feature fusion method can achieve better recognition performance while preserving lower feature dimensionality than the methods that uses single pedestrian feature and the simple fusion strategy of directly connecting multiple types of pedestrian features.