一种融合光流的分通道帧差目标检测方法  

An Object Detection Method Based on Channel Differencing Combined with Optical Flow

作　　者：夏瑜[1,2] 吴小俊[1] 冯振华[1] 

机构地区：[1]江南大学物联网工程学院,江苏无锡214122 [2]江苏广播电视大学常熟学院,江苏常熟215500

出　　处：《弹箭与制导学报》2012年第1期175-178,184,共5页Journal of Projectiles,Rockets,Missiles and Guidance

基　　金：教育部新世纪优秀人才计划项目(NCET-06-0487);国家自然科学基金(60973094;60572034;90820002;61070121);江苏省自然科学基金(BK2006081);江南大学创新团队研究计划项目(JNIRT0702)资助

摘　　要：针对目前方法在复杂环境下很难有效检测出运动目标的问题,提出了融合光流的分通道帧差目标检测方法。首先通过降噪和平滑预处理工作增强有效信息,然后使用实时分通道图像差分精确检测运动目标,最终融合光流检测信息,以运动信息修正检测误差,很好的逼近真实目标。在检测算法中使用数学形态学方法去除噪声和斑点以提高检测效果。实验结果证明,该算法能对运动目标快速而准确的检测。In view of moving object detection is satisfactory with traditional methods in complex environment,an object detection method based on channel differencing combined with optical flow was proposed in this paper.Firstly,pre-processing including noise reduction and smoothing was introduced to enhance the useful information.Then the moving object was accurately detected and extracted by real-time channel-difference.Lastly,the proposed method combined with optical flow can detect true target and correct detection error by movement information.To improve the effect of detection,mathematical morphology was incorporated into the proposed method to wipe off noise and strain.The experiments show that this method can detect moving objects in image sequence quickly and accurately.

关 键 词：运动目标检测 色彩通道 分通道帧差 数学形态学 光流 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]