光电成像制导中地面运动目标检测与跟踪  

Detection and tracking of moving targets in photoelectric imaging guidance

作　　者：李少军[1] 肖利平[1] 高磊[1] 

机构地区：[1]北京航天自动控制研究所,北京100854

出　　处：《系统工程与电子技术》2014年第1期23-30,共8页Systems Engineering and Electronics

基　　金：国防预研基金资助课题

摘　　要：地面运动目标的自动检测与跟踪是对地精确打击光电成像制导中的核心任务。针对地面运动目标自动检测问题,提出了由粗到精的迭代全局运动补偿算法以补偿动平台引起的背景图像像素位移,基于变化能量测度的独立运动目标检测算法以及基于三帧序列图像的运动目标定位算法。针对地面运动目标跟踪问题,提出了融合运动特征和灰度直方图特征的粒子滤波器目标跟踪算法,实现可靠、稳定跟踪地面运动目标。利用实际航拍视频进行算法验证实验,比较了基于傅里叶-梅林变换(Fourier-Mellin transform,FMT)、Kanade-Lucas-Tomasi(KLT)特征跟踪、Harris角点和迭代全局运动补偿算法,验证了基于变化能量测度的独立运动目标检测算法和融合多特征的粒子滤波器跟踪算法的有效性。The automatic detection and tracking of ground moving targets are main tasks in the photoelectric imaging guidance. For the detection of ground motion, a coarse to fine iterative global motion compensation method is proposed to compensate the displacement of image pixels due to the motion of sensor, and a moving target detection algorithm based on the change energy measure is presented to extract the region of moving targets on each level of image pyramid and difference image pyramid. The moving targets are located based ontwo change energy images evaluated from the three sequential images. For the ground motion tracking, particle filters fused motion feature and gray histogram are proposed to reliably track the moving targets on the ground. In the practical aerial video, the iterative global motion compensation method is compared with Fourier-Mellin transform （FMT）, Kanade-Lucas Tomasi （KLT） feature tracker and Harris corner, and the detection algorithm with change energy measure and the tracking algorithm with particle filters are validated.

关 键 词：光电成像制导 地面运动目标自动检测与跟踪 迭代全局运动补偿 变化能量测度 粒子滤波器 特征融合 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]