基于相关滤波器的目标跟踪方法综述

摘要

目标跟踪是计算机视觉中的重要组成部分,广泛应用于军事、医学、安防、自动驾驶等领域.虽然取得了很大进展,但由于遮挡、快速运动、模糊、光照及变形等因素存在,其仍是具有挑战性的研究领域.近年来,属于判别式类型的相关滤波器跟踪方法由于具有非常高的处理速度备受关注.首先介绍了目标跟踪和相关滤波器的基本知识,之后对相关滤波器方法在朴素阶段、循环结构和核技巧、多特征通道、与深度特征的结合、尺度研究、边界效应以及其他信息的利用方面进行了详述,最后对基于相关滤波器方法的研究方向和发展趋势给出了几点看法.%Object tracking is an important part in computer vision and is widely used in military,medical,security and autonomous driving.Although great progress has been made,it is still a challenging research field due to the factors such as occlusion,rapid speed,motion blur,illumination and deformation.In recent years,the correlation filter tracking method,one of discriminant type,has attracted much attention due to its higher processing speed.We first introduces the basic knowledge of the object tracking and the correlation filter tracking,and the correlation filter tracking methods in simple stage,we also discussed the circular structure and the kernel trick,the combination of multiple feature channels and deep feature,scale research,boundary effect and the use of other information.Finally,the research direction and development trend of the method based on the correlation filter is given.