相机运动状态下的实时运动目标跟踪算法  

Real-time Object Tracking Algorithm with Cameras Mounted on Moving Platforms

作　　者：姜明新[1] 

机构地区：[1]大连民族学院信息与通信工程学院,辽宁大连116605

出　　处：《大连民族学院学报》2014年第1期74-77,87,共5页Journal of Dalian Nationalities University

基　　金：中央高校基本科研业务费专向资金资助项目(DC110321);辽宁省教育厅科学研究项目(L2012476);大连民族学院2013年精品实验项目

摘　　要：提出了一种相机运动状态下的实时运动目标跟踪算法。根据运动目标对光流结果的影响对特征点进行筛选;利用光流法对相机的全局运动进行估计;根据全局运动估计的结果对粒子滤波的运动方程进行修正,选取颜色直方图作为目标的特征模型,实现对移动目标的跟踪。实验结果表明,在相机运动的状态下,能够准确快速跟踪运动目标,可以达到实时性要求,具有非常好的鲁棒性。In this paper proposed a real -time object tracking algorithm with cameras mounted on mobile platforms. Firstly, we rule out feature points that have inconsistent optical flows with those of background are ruled out. Secondly, optical flows on the rest feature points are used to estimate the global motion of camera. Finally, the kinematic function of particle filtering is mod- ified by the global motion of camera, together with color - space histogram as appearance model, to achieve robust tracking performance in unstable video sequences. The experimental results demonstrate that our method achieves promising real - time tracking performance and can track moving object robustly in videos captured by moving cameras.

关 键 词：移动平台 特征点筛选 全局运动估计 粒子滤波算法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]