基于对比度分析及矢量匹配的混合高斯模型

Gaussian mixture background model based on contrast analysis and vector matching

针对经典的混合高斯背景建模算法鲁棒性不强且背景建模实时性不足的特点,提出了一种改进方法。首先将图像矢量化,即将图像分成若干块,每一块图像作为一个矢量进行整体建模;然后对于每一个图像块基于其反差描述元与K个高斯模型进行匹配。实验结果表明,改进的算法降低了环境光变化和背景波动等因素的干扰且建模速度较快。

This paper introduces an improved methods on Gaussian mixture background model to enhance its robustness and shorten the modeling time. It divides an image into blocks at first, and build overall modeling for each block as a vector. Then matching every vector to K Gaussian models based on its descriptor. Experimental resuhs verify that the improved model reduces the interference as ambient light change, background fluctuations as well as time consumption.