一种基于多层背景模型的前景检测算法

动态场景中的前景检测是后继处理的基础和制约整个智能视频监控系统稳定性、可靠性的关键。为了在保证运动目标检测的基础上,进一步检测出前景中的静止目标并消除"鬼影(Ghost)",提出了一种基于多层背景模型的前景检测算法。该算法将背景分为参考背景和动态背景两层,分别采用单高斯和混合高斯模型进行背景建模。在线检测时,采用动态背景提取变化前景,用动态背景与参考背景之间高斯分布的差异提取静止前景,同时,通过逐层分析,比较输入像素与两层背景模型分布的相互关系,快速消除Ghost,降低虚警。实验结果表明,多层背景模型具有良好的检测性能和实时性,为后继跟踪、分类等处理提供了坚实的基础。目前,以该算法为核心构建了一个实时目标检测、跟踪系统,对图像大小为320×240的视频序列的平均处理速度达到15帧/s。

A Multiple Layer Background Model for Foreground Detection

Foreground detection is an important research problem in visual surveillance. In this paper, we present a novel multiple layer background model to detect and classify foreground into three classes, moving object, stationary object and ghost. The background is divided into two layers, reference background and dynamic background. Single Gaussian model and Gaussian mixture model are used respectively. Compared with many existing background models, an unique characteristic of the proposed algorithm is that through analyzing the Gaussian distributions of the two layers, stationary object and ghost are correctly labeled. Real time object detection and tracking system is developed and tested under indoor and outdoor scenes with various scenarios. Extensive experimental results demonstrate that the proposed algorithm is effective and efficient and the processing speed of the system reaches 15fps for the image size of 320×240.