边界与区域相融合的非结构化道路检测算法

为了提高非结构化道路识别算法的有效性，提出在运用图像分类和使用归一化彩色分量进行判断的基础上，充分利用图像的边缘信息，通过将图像边界范围、边界走向趋势与区域分类的结果相融合，得到道路的边界。使用Kalman滤波器对道路进行跟踪，处理了由于石阶等原因造成的路面扩大、路与非路交界区模糊难以区分和道路受阴影影响等实际情况。实验结果表明，该方法简便有效，能适应各种不同的非结构化道路，具有一定的鲁棒性。

Measurement Arithmetic for Unstructured Road Based on Boundary and Region

In order to enhance the validity for the unstructured road recognition, this paper makes full use of the picture edge information after image classification and judgment of normalized color tristimulus coefficient to get the boundary data of the road through the image boundary scope, the boundary line tendency and the region classification. The following step should be tracking the path with Kalman Filter(KF), which solves the problem of road surface expansion, vagueness in road and un-road border area and shadow influence due to stone steps. The experimental result indicates that this simple and effective method adapts to every kind of different unstructured road and has certain robustness.