基于改进TLD的自动目标跟踪方法  

Automatic tracking method based on improved TLD

作　　者：易诗[1] 林凡强[1] 周姝颖[1] 

机构地区：[1]成都理工大学信息科学与技术学院,四川成都610059

出　　处：《重庆邮电大学学报（自然科学版）》2016年第6期892-896,共5页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

摘　　要：视觉跟踪一直是机器视觉研究热点,TLD(tracking-learning-detection)算法是近年来出现的一种高效的视觉跟踪算法,针对TLD算法中Lucas-Kanade(LK)光流法无法有效跟踪物体快速移动和尺度变化的问题,采用金字塔光流法对TLD算法进行改进。并将所跟踪物体形心作为图像定位参考点,提取物体定位信息,通过定位信息运用比例-积分-微分(proportion-integral-derivative,PID)控制算法控制摄像头舵机云台转向,使摄像头快速、灵活、精确地自动跟踪指定物体。通过系统测试,与传统TLD算法对比,采用金字塔光流法改进的TLD目标跟踪算法在跟踪目标发生光照变化、尺度变化等情况时,具有更加优良的跟踪性能,准确将跟踪目标形心位置提供给控制部分,控制算法高效灵活,在获取信息后精确、快速地控制摄像头方位,使其正对跟踪目标。该系统对目标跟踪技术、安防技术、自动瞄准系统具有重大意义。Visual tracking has been a hot research topic in machine vision,and tracking-learning-detection（ TLD） algorithm is a kind of efficient visual tracking algorithm emerging in recent years. Aimed at solving the problem that the Lucas Kanade（ LK） optical flow method can not effectively track the object＇s rapid moves and scale changes,this paper uses the pyramid optical flow method to improve the TLD algorithm. Tracking the center of the object as the reference point for the image positioning to extract the object location information,using the positioning information and the algorithm PID control algorithm to control the camera gear steering,this way camera can rapidly,flexibly,precisely and automatically track the specified object. By testing the system,improved TLD target tracking algorithm using pyramid optical flow method under illumination and changes has more excellent tracking performance,compared with the traditional TLD algorithm for target tracking. The flexible algorithm can provide accurate target center position for the controlled part,it can steer camera accurately and fast to face the target directly after accessing the information. The system is of great significance to the target tracking technology,security technology,and automatic targeting system.

关 键 词：TLD算法 金字塔光流法 图像定位 比例-积分-微分(PID)控制算法 

分 类 号：TN919.5[电子电信—通信与信息系统]