基于几何信息增强的乡村道路自动提取方法

An Automatic Extraction Method of Rural Road Based on Geometric Information Enhancement

随着无人机及卫星影像技术的发展,乡村道路的自动提取对于大面积土地利用调查工作具有重要意义,能够显著加快调查的效率,减少测绘人员的工作量。然而,乡村遥感影像中存在较多的不可渗透面及裸地,这为自动提取道路带来了同谱异物的干扰现象。基于面向对象地物提取方法的基本原理,利用乡村道路的分割对象具有显著几何特征(长宽比、紧致度等)的特点,对其他不可渗透面及裸地进行有效区分,从而提高乡村道路自动提取的精度。通过南京市江宁区某乡村的无人机影像进行提取实验,本文方法具有较好的识别精度,准确率为92.8%,具有较好的实用性。

With the development of UAV and satellite image technology,automatic extraction of rural roads is of great significance for large-area land use survey,which can significantly improve the efficiency of survey and reduce the workload of surveying and mapping personnel.However,there are many impermeable surfaces and bare land in rural remote sensing images,which brings the interference of homospectral foreign matter to the automatic extraction of roads.Based on the basic principle of object-oriented feature extraction method,the segmentation object of rural road has significant geometric characteristics(length width ratio,compactness,etc.)to effectively distinguish other impermeable surfaces and bare land,so as to improve the accuracy of automatic extraction of rural road.Through the extraction experiment of UAV image in a village of Jiangning district,Nanjing,this method has good recognition accuracy,the accuracy is 92.8%,and has good practicability.