运动目标检测和跟踪算法综述

A summary of Moving Target Detection and Tracking Algorithm

运动目标检测和跟踪在多媒体图像、视频监控等方面有普遍应用。多年来,人们在该领域进行了大量的、深入的研究,发表了大量显著性的成果。但目标被遮挡、尺度变化以及光照变化仍然对跟踪结果存在较大影响。为解决这些问题,研究人员仍然在研究如何能构造一个鲁棒性好的跟踪算法。该文主要对近年来常见的运动目标检测和跟踪算法的发展做了简单叙述。简单分析了运动目标检测的帧间差分法、背景减法、光流法的原理;描述了运动目标跟踪所用的Meanshift算法、Kalman滤波器、粒子滤波方法;最后对压缩感知理论进行简介,并对基于该理论的两种算法:稀疏表示的目标跟踪算法和实时压缩感知目标跟踪算法做出系统的描述。

Moving target detection and tracking are widely used in multimedia images, video surveillance,etc. Over the years, peoplehave done a lot of in-depth research in this field, and published a lot of significant results. However, target occlusion, scale changeand illumination change still have a great impact on tracking results. To solve these problems, researchers are still studying how toconstruct a robust tracking algorithm. In this paper, we briefly describe the development of common algorithms for moving objectdetection and tracking in recent years. The principle of frame difference method, background subtraction and optical flow methodfor moving object detection is simply analyzed. The Meanshift algorithm, Kalman filter and particle filter method for moving targettracking are described. In the end, a brief introduction to the theory of compressed sensing is made. And two algorithms based onthis theory are described: sparse representation tracking and compressive tracking.