基于机器视觉的丘陵山区田间道路虚拟中线提取方法
A Machine Vision-Based Method for Detecting Virtual Midline of Field Roads in the Hilly Areas

田间道路识别是农业机械在田间道路上自动行驶的基础.针对丘陵山区田间道路复杂多变、无车道线和无明显边界等特点，提出一种基于机器视觉的道路虚拟中线提取算法.首先将道路RGB图像转换到HSI空间，选择与图像彩色信息无关的I分量；然后利用二维Otsu阈值分割法提取道路区域特征，去噪处理后选取目标区域分块求取质心点，对求取的质心点进行基于曲率变化的拟合，得到田间道路的虚拟中线.试验结果表明，在光照、水渍等不利因素影响较小的情况下，该算法拟合确定的道路中线与实际的道路中线相差最大不超过5%，准确性较高，能够有效实现丘陵山区田间道路的识别和虚拟中线的提取.
The recognition of field roads is the foundation of automatic driving for agricultural machinery. Taking into consideration the fact that the field roads in hilly areas are complex and have no distinctive edges, the authors of this paper propose an algorithm for extracting the virtual midline of roads based on machine vision. Firstly, the RGB image of the road is converted into the HSI space, and the component I, which is independent of the color information in the image, is selected. Then, the road feature is extracted by the 2D Otsu's thresholding method of segmentation. After noise point removal in the image, the target area is divided into several parts to seek their centroid points. Finally, the centroid points are fitted by curvature variation, and the virtual midline of the field road is obtained. Simulation results show that the difference between virtual midline and actual midline of the road is no more than 5%, indicating a high reliability. The algorithm can identify the road area effectively and detect the virtual midline of roads accurately under the conditions that illumination, water stain and other unfavorable factors have no significant influences