基于视频图像的目标检测与跟踪算法研究

发布时间：2018-07-31 07:03

【摘要】：随着科学技术的不断发展,计算机视觉成为一项热门的研究内容,目标检测与跟踪技术是计算机视觉的一个重要的研究方向。目标检测与跟踪是计算机通过相关算法,从视频图像序列中检测出感兴趣的目标物体,然后确定目标的位置信息,并且对目标物体的形状、速度和运动轨迹等进行描述。目前,目标检测与跟踪算法已经广泛地应用于我们的生活。但是,由于受到背景环境、设备质量和光照变化等多种因素的影响,目前的目标检测与跟踪算法几乎都有一定的局限性,在特殊的环境下对目标的检测与跟踪的效果很差,甚至无法完成对目标的跟踪任务。因此,设计一个稳定性好、复杂度低、准确性高和适应性强的目标检测与跟踪算法依然是一个重要且艰难的研究方向。本文分析并总结了国内外的研究现状,在此基础上对视频图像的单目标检测跟踪算法进行了研究。传统的目标检测跟踪算法实时性好,利于跟踪,但是当目标在运动过程中发生尺度变化时,将得到很差的跟踪结果,不仅对目标的尺寸变化敏感,而且缺乏对模板的实时更新。针对这个问题,对传统的目标检测跟踪算法中的缺点和不足加以改进,以求得到更好的跟踪效果,本文指出了距离加权和控制因子对目标检测与跟踪的影响,并改进了传统的基于Mean-Shift算法的目标跟踪算法,通过提取分析比较目标与模板的边缘加权直方图和中心加权直方图的相似度,可以有效的反映出目标的尺寸变化情况,并以此作为依据实现核函数带宽的自适应。
[Abstract]:With the development of science and technology, computer vision has become a hot research content. Target detection and tracking technology is an important research direction of computer vision. Target detection and tracking is that the computer detects the object of interest from the video image sequence through the correlation algorithm, then determines the position information of the target, and describes the shape, velocity and trajectory of the object. At present, target detection and tracking algorithms have been widely used in our life. However, due to the influence of background environment, equipment quality and illumination change, the current target detection and tracking algorithms almost have some limitations, and the effect of target detection and tracking is very poor in the special environment. You can't even track the target. Therefore, it is still an important and difficult research direction to design a target detection and tracking algorithm with good stability, low complexity, high accuracy and strong adaptability. This paper analyzes and summarizes the current research situation at home and abroad, and then studies the single target detection and tracking algorithm of video image. The traditional target detection and tracking algorithm has good real-time performance and is beneficial to tracking. However, when the target changes in scale during the moving process, it will get poor tracking results, which is not only sensitive to the change of target size, but also lack of real-time updating of template. To solve this problem, the shortcomings and shortcomings of traditional target detection and tracking algorithms are improved to obtain better tracking results. The influence of distance weighting and control factors on target detection and tracking is pointed out in this paper. The traditional target tracking algorithm based on Mean-Shift algorithm is improved. By extracting and analyzing the similarity between the edge weighted histogram and the center weighted histogram of the target and the template, the size change of the target can be effectively reflected. Based on this, the kernel function bandwidth adaptation is realized.
【学位授予单位】：长江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】