基于HVS的视频监控目标提取方法研究  

HVS-based object detection algorithm for surveillance video

作　　者：赵潇 刘畅 田霖[2,3] 韩雪 周继华 吴坚[5] Zhao Xiao;Liu Chang;Tian Lin;Han Xue;Zhou Jihua;Wu Jian(School of Communications & Information Engineering, Chongqing University of Posts & Telecommunications, Chongqing 400065, China;Beijing Key Laboratory of Mobile Computing & Pervasive Device, Beijing 100190, China;Wireless Communication Technology Research Center, Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100190, China;University of Chinese Academy of Sciences, Beijing 100190, China;Chongqing Jinmei Communication Co, Ltd, Chongqing 400030, China)

机构地区：[1]重庆邮电大学通信与信息工程学院,重庆400065 [2]移动计算与新型终端北京市重点实验室,北京100190 [3]中国科学院计算技术研究所无线通信技术研究中心,北京100190 [4]中国科学院大学,北京100190 [5]重庆金美通信有限责任公司,重庆400030

出　　处：《计算机应用研究》2019年第7期2214-2218,共5页Application Research of Computers

基　　金：北京市自然科学基金资助项目(L172049);北京市青年拔尖人才资助项目(2015000021223ZK31)

摘　　要：目标识别是实现视频监控智能分析的基础,但在光照、阴影以及杂乱背景等场景中,往往会出现目标误判以及不合理聚类等问题。针对上述问题,提出一种基于人类视觉系统(HVS)的视频监控目标提取方法。结合HVS视觉关注原理,优化背景差法检测结果中存在的重复检测和错误分割问题,并根据HVS的跟踪特点以及目标运动的连续性,结合相邻帧检测结果以达到目标区域的完整准确提取。最后,基于实际采集视频进行仿真实验,证明所提目标检测算法结果准确性更高,在复杂背景下也有良好的检测效果。Object detection is the basis of intelligent analysis, however, in the scene of illumination, shadow and clutter background, the problems of object misjudgment and unreasonable clustering are often appeared. Aiming at the above problems, this paper proposed a HVS-based object detection algorithm, which could optimize the error judgment and segmentation. According to the tracking characteristic of HVS and the continuity of object movement, the algorithm combined the detection results of adjacent frames to extract object area completely and accurately. Finally, the simulation experiment based on the actual acquisition videos shows that the proposed algorithm is more accurate and has good effect in complex background.

关 键 词：监控系统 目标提取 背景差法 人类视觉系统 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]