基于Mean Shift的视频图像检测与跟踪

在复杂背景的视频图像中,实时、准确、连续、长距离的跟踪以人为对象的目标,是一件很困难的任务。人体对象在跟踪目标图像位置的变化时,一直随着姿态的变化而改变,因此这是一个非常典型的非刚体目标,对这类目标采用简单的模板匹配的方法进行目标跟踪,无法达到准确的跟踪。均值漂移（Mean Shift）是现今最受欢迎的对象跟踪方法之一,广泛的运用于人脸的跟踪,文章提出了一种基于均值漂移算法的复杂背景视频图像检测与跟踪算法。在运动目标跟踪中,提出了以直方图为模式特征,以均值漂移算法为核心算法的目标跟踪算法,通过实验表明该跟踪算法能对候选目标进行运动检测,完成实时跟踪,同时有效抑制了局部遮挡、背景混乱等,过滤了伪目标,保证了跟踪的可靠性。

Video image detection and tracking via Mean Shift algorithm

Realizing a real time,accurate,continuous and long-distance track detection and tracking in a video image with complex background is a very difficult task.When tracking the position of the target image,the human target keeps changing according to its posture,so this is a very typical nonrigid target.Therefore,to this kind of target and using simple template matching method for target tracking cannot achieve accurate tracking.Nowadays,Mean Shift is one of the most popular target tracking methods,and is widely used in human face tracking.In this paper,we proposed a video image detection and tracking algorithm based on the MeanShift algorithm.A Histogram mode feature extraction via MeanShift is proposed for target tracking.Some experiments are taken to investigate the performance of the proposed method,and the result shows that the algorithm can achieve the real-time tracking,and effectively inhibit a partial occlulation,background confusion,and filtering the pseudo-target to ensure the tracking reliability.