基于RGB熵的非结构化道路分割方法

摘要

根据非结构化道路彩色图像中,道路通常呈现为灰色,非道路区域呈现为彩色的特征,提出一种基于RGB熵的边缘到边缘非结构化道路分割方法.计算RGB熵直方图的最小差值,选取RGB熵的阈值,初步检测出道路与非道路区域;将初步检测结果图像划分成多个方格,计算每个方格内黑白像素比例,转换成二值图像;采用轮廓跟踪法搜索道路左右边缘方格,采用边缘到边缘的方法,准确分割道路区域和非道路区域.实验结果表明,该方法能较好分割非结构化道路,抗干扰能力强,对道路形状不敏感,计算量较小,符合无人驾驶系统实时性要求.%In color image of unstructured road,road is usually gray,while non-road area is colored.According to this feature,a side-to-side method for road segmentation based on RGB entropy was proposed.The RGB entropy threshold was selected by counting the RGB entropy histogram's minimum difference,which was used to detect the road and non-road preliminarily.The result image was divided into many squares,and the proportions of black and white pixels in each square were calculated and converted to two values.The left and right boundary squares were searched using contour tracing method,and road area and nonroad area were segmented accurately using side-to-side method.Experimental results show that this method is effective in the segmentation of unstructured road,and is insensitive to the interference and the shape of road.At the same time,the method has small amount of calculation,so it meets real-time requirements for the vehicle automatic driving system.