一种车辆视觉导航中自适应背景模型提取算法

An Adaptive Background Model Extraction Algorithm for Vehicle Vision Navigation

为了解决统计背景模型在图像远景及车流量较大时容易出现错误像素点的问题,提出一种自适应背景模型提取算法。在车流量少时,利用统计直方图法提取背景;在车流量大时,利用帧差法获取图像中背景像素点,然后求其平均灰度值来重组的图像帧,最后利用统计直方图法进行背景建模。定义了算法比较标准,并据此比较了该算法与传统算法在不同车流量下的性能优劣。实验结果表明,通过分析图像中运动车辆像素信息区分车流量大小,实现对背景提取算法的自适应选择,使其在实际应用中更具针对性和灵活性,为准确地检测出运动车辆提供必要的基础。

In order to slove the problem that wrong pixels may occur at the time of the distant things and heavy traffic load while using the statistical background model,a self-adaptive background model extraction algorithm is proposed.At the time of light traffic,the background is extracted by statistical histogram.At the time of heavy traffic,the background pixels can be obtained via frame differencing.Then the background average value is calculated to update the image frames.Finally,the background model is established through the statistical histogram method.This algorithm is compared with the traditional algorithm according to the determinant criteria defined.The result shows that traffic volume can be distinguished through the information of image pixels.Since the method can find background automatically,it is well-targeted and flexible,and providing the necessary basis for accurately detecting the moving vehicles.