三帧差结合改进高斯建模的运动目标检测算法

Moving target detection algorithm based on improved Gaussian mixture modeling and three-frame differencing

针对混合高斯模型对光照突变比较敏感以及当运动物体速度较慢时容易产生"鬼影"现象,提出了一种动态自适应学习率的高斯混合模型。通过融入帧差法将每帧的图像分为已运动区域、正在运动区域以及背景区域,分别给予不同的更新率来更新高斯混合模型。为了能够适应光照或者背景突变的情况,背景区域给予动态更新率,并且给予高斯模型更快速的更新策略,使用高斯混合模型与三帧差法相结合。实验结果表明,该算法有效的处理了"鬼影"、阴影现象以及建模速度的问题,具有很好的实时性以及抗干扰能力,能够精确的检测出运动目标。

Aiming at the difficulties of the illumination change and ＂ghost＂ phenomenon caused by the slow speed of moving tar- get. The dynamic adaptive learning rate Gaussian mixture modeling is proposed, the image of each frame is divided into moving area, moved area and background area through the integration of temporal differencing. Different parts are given different lear- ning rates to update the Gaussian mixture model. The background region is given dynamic learning rate to adapt the illumination mutational status. The background Gaussian model is also given more quickly update strategy. At last, Gaussian mixture mode- ling is combined with three-frame-differencing. The final experimental result shows that this algorithm effectively deals with the ＂ghost＂ and shadow phenomenon and the problem of modeling speed, which has good real-time performance and robustness, and can detect the moving targets accurately.