航拍图像序列的多运动目标定位

A Moving-Object Location System of Aerial Video

针对航拍图像序列中运动目标使用仿射模型来描述航拍图像序列中的背景运动 ,提出了一种基于稀疏采样的局部补偿误差函数泰勒展开算法 ,该算法能有效地估计出模型参数 ;又根据图象序列中背景运动与目标运动的矢量不一致的特点 ,采用矢量聚类的算法 ,实现了对多个目标的准确定位 .提出的算法对噪声和光照都具有强抑制性 ,可以在图像质量相对较差时 ,能确保对运动目标的快速定位 .

This paper realized a location system which locates moving-objects in aerial video. A sparse sample-based Taylor expansion method of local compensation error function to estimate the affine transform model parameters effectively was presented. Also a motion-vector clustering algorithm was presented to locate the objects in the video precisely. The methods presented above tolerate heavy noise and light- effects , and are fit for the situation when the quality of the pictures is relatively poor.