一种基于同态滤波的背景建模方法

摘要

在道路视频检测技术中提出了一种基于同态滤波的背景建模法.对图像进行同态滤波的预处理,以适应渐变的光照情况.通过图像分块与帧间差分来获取临时背景序列,为背景建模做准备.为了处理背景中由于车辆突然开动或停止产生的"鬼影",对帧序列有效区中的每一像素进行基于灰度分区直方图模式建立背景模型.这种方法既能应对光照渐变又能适应大流量且车速慢的拥堵情况.实验结果表明,此算法具有良好的可靠性与鲁棒性.%This paper presents a homomorphic filtering-based background modeling approach. In coarse segment phase, the establishment background can be divided into three steps: adapting to the change of illumination condition using the image preprocessing by homomorphic filter, obtaining temporary background sequences after image region and frame difference to the preprocessed image, which is the provision of building background, and effective framing per pixel based on gray histogram partition background model to handle the "ghost" (shadow) because of the vehicles＇ sudden start or stop. The method is not only able to deal with the illumination change but also successfully tested in busy traffic and other complex scenes. Experimental results show that this algorithm has good robustness and reliability.