Abstract:

Aiming to solve the problem that traditional electronic image stabilization algorithm could not eliminate video random jitters  quickly and efficiently, based on subzone gray projection a  video stabilization method  was used to ensure the robot system outputting satellite assembly pictures stably and coherently. Two frames were divided and subregions of low contrast were deleted. The interval projection and cross-correlation algorithm were used to get local motion component. The global motion vector was solved by the rest local motion component after iterative steps based on the average error threshold. An average filter would be used in global motion vector of multiple frames if image had low frequency scanning component. Experimental  results show that using the stabilization method based on subzone gray projection's  image the  stabilizing precision  increases 119.1%  in natural scene images and 55.8% in simulation  assembly images of low contrast compared with traditional gray projection method, meanwhile the execution time is only 0.5% of block matching algorithm. It proves that the method can eliminate random jitters effectively, output stable coherent video images rapidly, and guarantee the satellite installation of robot system.

Key words: satellite , assembly;electronic image stabilization;subzone gray projection;random jitter

摘要：

针对传统电子稳像算法无法快速有效地消除视频图像随机抖动的问题,采用一种基于分区灰度投影的稳像算法，以确保机器人系统能够输出稳定连贯的卫星装配画面。对前后两帧视频图像进行划分并删除对比度低的子区域，利用间隔投影和互相关运算获取局部运动分量,通过基于平均误差门限的迭代步骤筛选后剩余的局部运动分量求解全局运动矢量；若判定存在低频扫描分量，还需对多帧图像的全局运动矢量作均值滤波处理。实验结果表明：基于分区灰度投影的稳像算法相比传统灰度投影法，在低对比度的自然场景图像和模拟装配图像中的稳像精确度分别提升119.1%和55.8%;同时执行时间只有块匹配算法的0.5%。能够有效消除随机抖动,快速输出稳定连贯的视频画面，保证机器人系统顺利完成卫星的地面装配工作。

关键词: 卫星装配, 电子稳像, 分区灰度投影, 随机抖动