地面剧烈旋转运动背景中运动目标检测  

Moving object detection in violently rotational ground background

作　　者：闫钧华[1,2] 储林臻[1] 艾淑芳 李大雷 段贺[1] 

机构地区：[1]南京航空航天大学航天学院,南京210016 [2]光电控制技术重点实验室,洛阳471009

出　　处：《仪器仪表学报》2014年第9期2078-2086,共9页Chinese Journal of Scientific Instrument

基　　金：光电控制技术重点实验室和航空科学基金联合项目(20135152049);航天科技创新基金项目资助

摘　　要：针对地面背景剧烈旋转运动下运动目标难以检测的问题,提出基于彩色光流计算与运动补偿的运动目标检测算法。首先利用金字塔分层计算、Lucas局部运动一致性,提出一种改进的彩色金字塔Lucas光流算法,以获得背景剧烈旋转运动下的准确可靠的图像光流场;然后基于光流场估计背景运动透视模型参数,重构图像纯背景运动矢量场,对图像进行背景运动补偿;最后通过差分二值化图像与形态学处理检测出运动目标。实验结果表明算法能够有效地改善光流计算效果,光流场局部方向一致性平均误差为0.29°,局部矢量模值一致性平均误差为0.25pix,该算法能在地面剧烈旋转运动背景中准确地检测出运动目标。Aiming at the difficulty in moving object detection under violently rotational ground background,we propose a moving object detection method based on color optical flow field and motion compensation.First,an improved color pyramid Lucas optical flow algorithm is proposed,which is based on image pyramid calculation and Lucas local motion coherence.With the help of the improved color optical flow algorithm,the accurate,reliable optical flow field is obtained,which reflects the movement of the image pixels under violently rotational ground background.From the optical flow field the parameters of the perspective background motion model are estimated and the motion vector field for pure background information is reconstructed.Then the background difference of two frames can be removed by motion compensation.Finally,the moving object is identified through the frame difference method and morphological processing.Experiment results show that the method can significantly improve the quality of optical flow field calculation ; the average error of local direction consistency of the optical flow field is 0.29 degree ; the average error of local vector modulus consistency of the optical flow field is 0.25 pixel.The proposed algorithm can accurately detect the moving object from violently rotational ground background.

关 键 词：目标检测 彩色光流场 地面背景 旋转运动 运动补偿 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TH7[自动化与计算机技术—计算机科学与技术]