中文摘要：

提出了一种通过建立不同视角下目标间的关联度函数进行目标再确认的方法,用关联度值的高低来衡量目标间的同一性程度,并用于跨摄像头目标再确认.该方法首先提取目标整体和局部特征的直方图,然后对每种直方图分别计算推土机距离(EMD),并对不同的特征分量分配不同的加权值,建立不同视角下目标的关联度函数.通过对不同摄像头下的多组目标进行再确认实验,并与传统的基于余弦角距离度量算法相比较,结果表明该算法取得了较好的目标再确认结果,具有一定的优越性.

英文摘要：

A correlative degree of targets between different camera views was established to do target re-identification.This method uses the value of correlative degree to measure the identity of targets. Firstly,we need to extract the global and local features of targets,and then calculate the earth mover′s distance (EMD)of each histograms,respectively.Further more,different feature vectors were given different weights to establish the correlative degree function of different camera view.Ex-periments of targets re-identification for different cameras show that the proposed method can achieve good results.And compared with the traditional cosine angle distance metric,our method has certain advantages.