一种面向移动终端的混合跟踪定位算法  

Hybrid Tracking Algorithm for Mobile Terminals

作　　者：曹若琛 陈靖[2] 王涌天[1,2] 

机构地区：[1]北京理工大学计算机科学与技术学院,北京100081 [2]北京理工大学光电学院,北京100081

出　　处：《太原理工大学学报》2016年第4期506-512,共7页Journal of Taiyuan University of Technology

基　　金：国家高科技863计划项目:基于内容理解的高融合度增强现实技术(2013AA013802);国家自然科学基金资助项目:面向运动目标检测识别的高分辨率光学压缩成像理论与技术(61271375)

摘　　要：通过扩展卡尔曼滤波器(EKF)对手机移动终端的加速度计、陀螺仪以及视觉数据进行融合,实现混合跟踪。在此基础上,对基于EKF的数据融合算法进行改进,当失去视觉校正时,利用同步跟踪的背景区域速度更新目标区域速度,使算法在目标区域提取特征点数不足的情况下可以持续跟踪;同时引入参考坐标系,定义世界坐标系在参考坐标系中的姿态,校正在基于EKF的数据融合算法中由于光照遮挡等问题出现的角度跳变错误;最后,在实际环境中进行测试,实验结果显示本混合跟踪算法相比传统单一视觉的跟踪注册算法和以往基于EKF的数据融合算法具有明显优势。In the Mobile Augmented Reality field, accurate tracking is a challenging task. Hy brid tracking based on Visual and Multi-sensor generally provides more stable results than single visual tracking. This paper fuses the data from visual and multi-sensor using Extended Kahnan Filter （EKF）. On this basis, to improve the algorithm of data fusion based onEKF, awaytoup date the velocity of the target using background is presented, which makes tracking to work well when visible feature points are not enough;the reference coordinate system is introduced as well, the attitude of world coordinate system is defined in the reference coordinate system,and tile error caused by lighting, shelter and so on is corrected. Experimental results show tirol the mobile ter- minal hybrid tracking presented in the paper is advantageous.

关 键 词：移动增强现实 多传感器 混合跟踪 扩展卡尔曼滤波器 

分 类 号：TN912.34[电子电信—通信与信息系统]