视频中运动物体检测算法的研究

Research on Moving Object Detection Algorithm in the Video

为了从运动场景中(视频序列)提取高质量的运动物体,本文提出一种基于代数多重网格和帧间差分法相结合的运动目标检测算法,该方法利用代数多重网格对帧序列图像进行重构,分别得到重建的三层图像,再利用帧差法对处理过的相邻图像进行差分处理,从而提取场景中运动的物体,实验证明,此方法解决了帧间差分法不能提取对象完整区域和对场景噪声等敏感的问题,鲁棒性高,能准确的检测出在场景中运动的目标。

In order to extract high quality moving object from the movement in the scene( video sequence),this paper puts forward a kind of algorithm which based on algebraic multigrid and inter-frame difference method of moving object detection,the method refactor the images of frame sequence by using the technology of algebraic multigrid,then take the three independent layers of image reconstruction. The method also use frame differential method to deal with the processed images,and thus to extract the moving object in the scene. The experiments show that this method has solved the questions that inter-frame difference method cannot extract comprete object area and it makes sensitive to the noise of scene. The method also has high robustness,can accurately detect the target in motion in the scene.