基于多层深度特征融合的行人再识别研究

Person Re-Identification Based on Multi-Layer Deep Feature Fusion

本文采用多层深度特征融合方法,首先,利用经典卷积神经网络对行人图像进行处理;然后,对卷积神经网络每一层得到的特征进行PCA降维,保留其主要成分,并将各层降维后的特征进行融合;最后,基于欧氏距离判断待查询行人与图像库中各行人的相似性,得到再识别结果.实验结果表明,与已有的行人再识别方法相比,本文方法准确率更高.

In this paper,a multi-layer depth feature fusion method is adopted.Firstly,the classic convolutional neural network is used to process pedestrian images.Secondly,in order to preserve main components,the principal component analysis is utilized to reduce the dimensionality of features obtained by each layer of convolutional neural network,and then the features of each layer after dimensionality reduction are fused.Finally,the similarity between pedestrians to identify and pedestrians in the image database is determined based on the Euclidean distance,and the result of re-recognition is obtained.Experimental results show that the proposed method is more accurate and robust than the previous algorithms.