基于小波变换与多帧平均法融合的背景提取

摘要

利用虚拟检测圈法对道路交通视频进行处理,可以检测出道路上运动车辆的车速.而虚拟检测圈法首先需要得到准确的视频道路背景图像.针对交通道路视频,在背景提取算法仿真分析的基础上,利用小波变换,将提取到的背景图像进行一级小波分解,得到各分量图像.通过对水平分量、竖直分量及对角分量进行去噪处理后,再采用小波重构得到最终的背景图像.采用该方法所得背景图像能有效地减少主要背景提取算法中存在运动车辆轨迹的缺点.%Using virtual detection loop method to deal with road traffic video, the speed of moving vehicles can be detected. But the virtual detection loop method needs accurate background image of road video firstly. Based on simulation analysis on background extraction algorithms, we use wavelet transform to extract the background image in order to get each component image. Through denoising the horizontal component, the vertical component and the angular component, these components are used to reconstruceted the final background image by wavelet transform. This method can effectively reduce the shortcomings with moving vehicle trajectory in the reconstructed background image, which exist in those main background extraction methods.