基于空时多线索融合的超像素运动目标检测方法

Moving Object Detection Method Via Superpixels Based on Spatiotemporal Multi-cues Fusion

运动目标检测是计算机视觉领域极具挑战性的难题,该文针对这一问题提出一种基于空时多线索融合的超像素运动目标检测方法。首先利用简单线性迭代聚类算法将当前帧分割为超像素集合,根据帧间的像素级时变线索找到当前帧中包含运动信息的前景超像素子块;然后根据运动目标的一致性原则建立前一帧目标模型,结合目标空间线索进一步确定包含运动目标的检测窗口,将目标检测问题转化为目标分割问题,利用密集角点检测将目标从窗口中分割出来。在多个具有挑战性的公开视频序列上同几种流行检测算法的实验对比结果证明了所提算法的有效性和优越性。

Moving object detection is a challenging issue in computer vision. In this paper, a new detection method via superpixels is proposed based on spatiotemporal multi-cues fusion. First, the current frame is segmented into a set of superpixels using simple linear iterative clustering and the subblocks of foreground superpixels containing motion information are captured according to the time-varying cue of inter-frame pixel-level. Then, a target model of the previous frame, which is established on the basis of the consistency principle of motion target and space clues of a target, are combined to further determine the detection window including the moving object. Finally, the problem of object detection is converted to object segmentation and an object is divided from the detection window utilizing the dense corner detection. Experimental results using several challenging public video sequences show the effectiveness and superiority of the proposed method compared with other state-of-the-art detection approaches.