摘要
视频目标再识别涉及计算机视觉领域的运动目标检测、跟踪、图像处理、特征提取、特征匹配等.现提出一种基于前景检测、彩色区域边缘直方图(REH)的视频目标再识别算法.前景目标检测能有效消除背景像素产生的冗余特征,结合HSV空间中的彩色区域边缘直方图,增强了对目标的特征描述.实验在笔者建立数据集和3Dpes上取得了86.7%和51.5%的识别率,进一步提高了视频目标再识别的准确率.
Video object re-identification involves on moving object detection,tracking,image processing,feature extraction and feature matching,etc.A video object re-identification algorithm based on foreground detection and color region edge histogram（REH）was introduced.Combined with the region color edge histogram in HSV space,foreground object detection could eliminate the redundant features caused by background pixel effectively and enhanced the description of the characteristics of the target.The experiment achieved a re-identification rate of 84.5% and 46.7%in the data set build by author and public data 3Dpes,showed that the algorithm has certain application value and improves the accuracy rate of video object reidentification.
出处
《中国计量学院学报》
2016年第3期324-329,共6页
Journal of China Jiliang University
关键词
视频目标再识别
区域边缘直方图
前景目标检测
特征匹配
video object re-identification
region edge histogram
foreground object detection
feature matching
