%0 Journal Article
%T 一种基于时-空多尺度的运动目标检测方法<br>Space-Time Multiscale Based Moving Object Detection Method
%A 卢军
%A 王子梁
%A 朱佳磊
%J 西北工业大学学报
%D 2017
%X 针对复杂背景下多运动目标检测问题，提出了一种基于时-空多尺度的运动目标检测方法。该方法特点在于目标检测前先对待检测图像进行空间多尺度运动显著性分析，定位图像中运动目标所处的运动显著性区域；然后在显著性区域内进行时间多尺度帧间差分，通过寻找最优的差分间隔来实现不同运动速度目标的检测。实验结果表明新算法能够有效降低虚警和漏检，同时具有较强的适应性和鲁棒性。<br>For motion detection in temporal-spatial domain, a space-time multiscale based moving object detection method was proposed, which can carry out the multiscale in spatial and temporal respectively for motion detection. We achieve multiscale in spatial by using salience detection, and use multiscale Interval-Frame Space (IFS) to deal with the motion salience regions for the end of multiscale in temporal. This method was implemented in space multiscale which reflected the significance of movement, and also in time multiscale which reflected the difference among multiscale frames in the significant region of movement. At last, experimental results show that the proposed method can reduce false alarm and false detection probability. Meanwhile, the method has strong adaptation and robustness
%K 时-空多尺度
%K 运动显著性检测
%K 像素&lt
%K br&gt
%K motion detection
%K space-time multiscale motion salience detection
%K motion significance detection
%K pixels
%U http://journals.nwpu.edu.cn/xbgydxxb/CN/abstract/abstract6730.shtml