基于局部深度匹配的行人再识别

发布时间：2018-04-30 08:56

  本文选题：行人再识别 + 分块匹配　； 参考：《计算机应用研究》2017年04期

【摘要】：针对行人再识别精度低的难题进行研究,提出了一种新的基于分块匹配的行人再识别方法。首先,引入带人体结构信息的人体DPM对行人外观进行分割,得到的带语义信息的身体部件作为匹配识别的基本单元;其次,基于深度神经网络模型提取各部件的深度特征作为匹配依据;再次,基于余弦距离判断各身体部件与目标行人对应部件的相似性;最后,融合所有身体部件的识别结果得到最终的再识别结果。实验结果表明,跟已有方法相比,该方法具有更好的鲁棒性,在识别精度上有较明显的优势。
[Abstract]:A new method of pedestrian rerecognition based on block matching is proposed to solve the problem of low accuracy of pedestrian rerecognition. Firstly, the human body DPM with human structure information is introduced to segment the appearance of pedestrians, and the body parts with semantic information are used as the basic unit of matching recognition. Based on the depth neural network model to extract the depth features of each component as the matching basis; thirdly, based on the cosine distance to judge the similarity between the body parts and the target pedestrian corresponding parts; finally, Fusion of all body parts of the recognition results to obtain the final recognition results. The experimental results show that the proposed method is more robust than the existing methods and has obvious advantages in recognition accuracy.
【作者单位】： 国家数字交换系统工程技术研究中心;
【基金】：国家自然科学基金资助项目(61521003,61379151) 国家科技支撑计划资助项目(2014BAH30B01) 河南省杰出青年基金资助项目(144100510001)
【分类号】：TP391.41;TP183
，

本文编号：1823915