基于改进的混合高斯模型的背景建模方法

提出了一种基于改进的混合高斯模型的背景建模方法，克服了经典混合高斯模型方法计算量大和对长时间静止物体转为运动及光照突变较为敏感的缺点。首先，在经典混合高斯模型方法的基础上，引入了一种新的高斯分布个数的自适应选择策略，提高了建模效率。其次，分析了经典混合高斯模型方法对长时间静止物体转为运动及光照突变较为敏感的原因，采用了一种不同区域更新率的自适应选择策略，能够迅速响应场景的变化，有效地解决了大面积误检问题。通过在典型的场景下与经典混合高斯模型方法进行比较，验证了本文算法的有效性。

A background modeling method based on improved mixture Gaussian model

The typical mixture Gaussian model method is computability of high cost and less robust to the conditions of sudden moving of the motionless objects and the instant illumination changing. To solve this problem, a background modeling method based on improved mixture Gaussian model is presented in this paper. Firstly, to improve the efficiency of modeling, an adaptive selection strategy of the number of Gaussian distributions is proposed. Secondly, through of analyzing the cause why the typical mixture Gaussian model method is less robust to the conditions of sudden moving of the motionless objects and the instant illumination changing, present an adaptive selection strategy of the update rate for different regions to respond to the changing of scene and solve the large areas of false detection problem. Comparing with the typical mixture Gaussian model method on different image sequences containing targets of interest in typical environments. Experimental results demonstrate the effectiveness of the proposed method.