中文摘要：

低空飞行器监控平台在突发事件现场往往面临恶劣环境以及飞行器自身抖动影响监控质量的挑战，而现有的视频稳像技术无法很好地解决较大场景变化和显著景深影响的问题。该文针对六旋翼低空监测平台在复杂环境下应急视频监控稳像的难点，提出了一种基于运动分解估算和混合基准帧比照的视频稳像算法。基于运动分解估算的运动估计算法利用矩阵分解原理，将全局运动分解成帧间运动和前帧运动的合成，保证了场景快速变化条件下运动估计的准确性和时效性。混合基准帧比照方法运用基于阚值设定的形式，综合考虑计算误差和稳像连续性两方面的因素，寻找较为平衡的基准帧更新算法，有效地解决了视频景深影响造成的误差。实验结果表明：该算法提高了运动估计的准确率和计算效率，可满足六旋翼飞行器监控视频的稳像要求。

英文摘要：

Low-altitude aircraft video monitoring platforms in emer- gency conditions will be challenged by harsh environments and air- craft jitter. Current video stabilization algorithms cannot account for the effects of rapid scene changes and the influence of the depth of field. This paper presents a video stabilization algorithm based on motion factorization estimates and a hybrid reference frame for a six- rotor aircraft platform. The global motion is factored using matrix faetorization theory into two parts for the motion between two con- secutive frames and the motion compensation by the previous frame to guarantee the precision and timeliness of motion estimates for rap- id scene changes. Caleulational errors and the need for continuous stabilization can be handled by a hybrid-reference frame with thresh- olds in a balanced algorithm that finds the reference frame by solvingthe characteristic points matching error problem to correct for the in- fluence of the depth of field. Tests show the effectiveness and effi- ciency of this algorithm for motion estimation for video stabilization in six-rotor aircraft video monitoring systems.