一种基于惯量椭圆位姿补偿的扩展目标精确定位方法  

A Method of Extended Target Tracking Based on Inertia Ellipse and Pose Compensation

作　　者：赵汝进[1] 刘恩海[1] 张启衡[1] 赵连军[1,2] 

机构地区：[1]中国科学院光电技术研究所,四川成都610209 [2]中国科学院大学,北京100049

出　　处：《光学学报》2014年第3期182-187,共6页Acta Optica Sinica

基　　金：国家973计划(2014CB744200)

摘　　要：针对空间扩展目标成像纹理特征少,背景单一,以及目标旋转、平移、形变等运动特征所导致的目标定位跟踪困难的问题,研究了基于惯量椭圆位姿补偿的目标定位跟踪方法(TTBPC)。通过惯量椭圆测姿法获取目标准确俯仰角姿态,将目标位置、姿态补偿到统一的坐标系下,结合最近距离特征点匹配原则,最终实现对空间扩展目标的精确跟踪定位。通过仿真图像和实际图像对TTBPC方法的有效性进行了验证,平移目标定位误差小于0.26pixel,旋转目标定位误差小于0.28pixel,实际空间扩展目标定位误差小于0.3pixel,这表明该方法应用于空间扩展目标定位的有效性和准确性。In order to solve the problem of space extended target tracking on the condition that the background of image is simple, the texture of target is lack, and the revolving and translating target is distorted gradually, one method of target tracking based on the inertia ellipse and pose compensation is proposed. The inertia ellipse is used to measure the pose of target accurately. The pose and location are compensated in uniform reference frame, and the principle of the closest distance is adopted to feature point matching. It is approached that the extended target is tracked accurately. Furthermore, experiments involving emulating image and real image indicate that the proposed method is effective, the error of tracking transferred target is less than 0.26 pixel, the error of tracking revolving target is less than 0.28 pixel, and the error of tracking real extended space target is less than 0.30 pixel, which indicates that the proposed method is effective on tracking the extended space target.

关 键 词：图像处理 姿态测量 惯量椭圆 目标跟踪 特征点 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]