期刊文献+

基于组件超分辨率的多分辨率车辆重识别方法

Multi-Resolution Vehicle Re-Identification Method Based on Component Super-Resolution
原文传递
导出
摘要 在真实的道路监控场景中,考虑到监控摄像头与车辆的距离、监控摄像头不同的朝向和监控摄像头的高度等因素,捕捉到的车辆图片的分辨率差异较大,严重影响车辆重识别的性能。为此,提出基于对比学习的多分辨率特征提取方法,利用组件分割器提取车辆的组件特征,并通过对比学习的方法融合多分辨率输入图像的全局与组件特征,充分提取车辆整体边缘和细节信息。同时提出基于组件超分辨率的特征增强方法,对车辆图像进行组件分割并去除背景,将超分辨率应用到车辆组件上增强组件表征,可有效避免全局噪声干扰。与近年来先进的重识别方法开展了大量对比实验,在MLR-VeRi776和VRIC数据集上开展的实验结果验证了提出的多分辨率车辆重识别方法具备一定优势。 In a real road monitoring scenario,considering the distance between the surveillance camera and the vehicle,the different orientations of the surveillance camera,and the height of the surveillance camera,the resolution difference of the captured vehicle images are obvious,which seriously affects the performance of vehicle re-identification.To this end,a multi-resolution feature extraction method based on contrastive learning is proposed.The component segmenter is used to extract the component features of the vehicle,and the global and component features of the multi-resolution input image are fused by the method of contrastive learning.At the same time,a feature enhancement method based on component super-resolution is proposed to segment the vehicle image and remove the background and apply super-resolution to vehicle components to enhance the component representation,which can effectively avoid global noise interference.A large number of comparative experiments have been carried out with advanced re-identification methods in recent years.The experimental results carried out on the MLR-VeRi776and VRIC datasets verify that the proposed multi-resolution vehicle re-identification method has certain excellence.
作者 黄文心 钟忺 张军 巫世峰 陈淑琴 李琳 刘文璇 HUANG Wen-xin;ZHONG Xian;ZHANG Jun;WU Shi-feng;CHEN Shu-qin;LI Lin;LIU Wen-xuan(School of Computer Science and Information Engineering,Hubei University,Wuhan 430062,China;School of Computer Science and Artificial Intelligence,Wuhan University of Technology,Wuhan 430070,China;ZhongQianLiYuan Engineering Consulting Co Ltd,Wuhan 430070,China;College of Computer,Hubei University of Education,Wuhan 430205,China)
出处 《武汉理工大学学报》 CAS 2022年第11期96-104,共9页 Journal of Wuhan University of Technology
基金 湖北省自然科学基金(2021CFB281,2021CFB513,2021BAA030) 国家自然科学基金(62271361,62276196)
关键词 多分辨率车辆重识别 组件分割 对比学习 组件超分辨率 multi-resolution vehicle re-identification vehicle component segmentation contrastive learning component super-resolution
  • 引文网络
  • 相关文献

参考文献1

二级参考文献2

共引文献2

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部