慢运动背景下运动目标提取算法*

提出一种慢运动背景视频序列下基于帧间背景图像匹配的运动目标检测和提取算法。该算法首先使用仿射变换模型来描述慢运动背景图像的运动变化，并使用基于光流约束方法求解该仿射变换模型参数，实现了相邻帧间图像的背景匹配；其次，采用背景匹配后的两帧图像差进行目标检测，使用自适应二值化区分变化与未变化区域；最后，使用形态学等图像算法进行后处理提取运动目标。算法经实验证明，在背景慢运动情况下可以有效地提取出运动目标。

Moving objects detection under slow moving background

This paper proposed a detection algorithm of moving objects based on background matching under a slow moving background. Firstly, used an affine transformation model to describe the background image movement. Secondly, gave a method based on optical-flow restriction to solve this affine model parameters. Thirdly,used this affine model to match the two adjacent frame backgrounds. Finally, performed a subtraction between these two matched adjacent frame backgrounds for detection and extraction of the moving objects. The several experiments show that the proposed algorithm is efficient.