动态背景下空时特性均显著的运动目标检测  

Spatiotemporal salient moving object detection in dynamic background

作　　者：赵燕熙 尚振宏[1] 刘辉[1] 李润鑫[1] ZHAO Yanxi;SHANG Zhenhong;LIU Hui;LI Runxin(School of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650500, China)

机构地区：[1]昆明理工大学信息工程与自动化学院,昆明650500

出　　处：《计算机工程与应用》2017年第5期170-175,180,共7页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61462052);云南省自然科学基金(No.KKSY201403049)

摘　　要：从序列图像中提取变化区域是运动检测的主要作用,动态背景的干扰严重影响检测结果,使得有效性运动检测成为一项困难工作。受静态图像显著性检测启发,提出了一种新的运动目标检测方法,采用自底向上与自顶向下的视觉计算模型相结合的方式获取图像的空时显著性:先检测出视频序列中的空间显著性,在其基础上加入时间维度,利用改进的三帧差分算法获取具有运动目标的时间显著性,将显著性目标的检测视角由静态图像转换为空时性均显著的运动目标。实验和分析结果表明:新方法在摄像机晃动等动态背景中能较准确检测出空时均显著的运动目标,具有较高的鲁棒性。Moving objects detection in image sequence is a crucial process in lots of computer vision applications. Although many approaches were proposed in past decades, dynamic background caused by camera movement, illumination changes and etc. in real scenes makes it a challenging task, especially when detecting spatiotemporal salient moving objects which is more meaningful in later processing. Inspired by the saliency detection in still image, a new method based on bottom-up and top-down visual computing model to obtain spatiotemporal salient moving object detection is proposed. Firstly, the algorithm detects the spatial saliency of image sequence, then, time dimension is added by applyingthe optimized method of three-frame difference to detect temporal saliency of moving object. So, the perspective of detecting salient object is transferred from the saliency map in still image to spatiotemporal context. Experimental results and analysis show that the new algorithm is more robust than lots of algorithms with the ability to detect spatiotemporal salient moving object in dynamic background.

关 键 词：显著性 运动目标 检测 三帧差分 动态背景 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]