Abstract: An abandoned objects detection method based on double background is presented in this paper. Sliding average model and improved Gaussian Mixture Model(GMM) are used respectively to construct separate reference and dynamic background. Compared the two prospects of the two layers, static object and ghost are labeled. Ghost can be eliminates by using the histogram matching method with the stationary objects. Based on the foreground detection, Mean Shift algorithm and particle filter algorithm are used to tracking. Experiment results prove the validity of the method.

Key words: Gaussian Mixture Model(GMM), slide average algorithm, ghost, static foreground detection, target tracking, abandoned object detection

摘要： 提出一种基于双层背景的遗弃物检测方法。分别采用滑动平均算法和改进的高斯混合模型，对参考背景和动态背景进行建模，通过2个背景得到前景间的差异，以此提取静止前景，对检测到的静止物体进行直方图匹配以消除鬼影，在前景检测的基础上引入均值漂移算法和粒子滤波算法，处理物体间遮挡问题。实验结果证明了该方法的有效性。

关键词: 高斯混合模型, 滑动平均算法, 鬼影, 静止前景检测, 目标跟踪, 遗弃物检测

CLC Number: 

• TP391