基于背景估计的运动检测算法

Motion detection algorithm based on background estimated

为了有效检测出运动目标静止后的轮廓和抑制视频抖动给运动检测带来的影响,提出了基于背景估计的运动检测算法。采用带模板遮罩的背景更新策略来避免目标前景区域被更新,同时根据纹理相似性,对目标遮挡区域的背景进行估算,使得在光照变化下遮挡区域得到同步更新。通过视频稳定算法得到全局运动参数和局部运动参数,然后利用全局运动参数稳定视频,利用局部运动参数作为运动目标检测的输入,最后利用高斯模型来检测运动目标。通过对多段抖动视频做实验,结果验证了该算法的准确性和有效性。

In order to detect the contour of a static target and suppress the impact of Jitter video in motion detection,a motion detection algorithm based on background estimated is presented.On one hand,the background updates strategy with template masks are adopted to avoid updating the target foreground area.At the same time,the algorithmestimate target background according to the texture similarity for the sake of updating the sheltered area synchronously under the changes in the illumination conditions.On the other hand,the global motion parameters and partial motion parameters with video stabilization algorithm are gained.The global motion parameters is used to stabilize video and the partial motion parameters is treated as the input of the target detection module.And then,the algorithm detects targets with Gaussian model.Finally,by making experiments under several jitter video segments,the validity and efficiency of the presented algorithm are demonstrated.