基于高分辨率遥感图像的道路提取研究  

Research of Road Extraction from High Resolution Remote Sensing Images

作　　者：李亚州 池润昊 宋菲[2] 徐昇[1] LI Yazhou;CHI Runhao;SONG Fei;XU Sheng(College of Computer Science and Technology,Nanjing Forestry University,Nanjing 210037;School of Information Engineering,Jiangsu Open University,Nanjing 210017)

机构地区：[1]南京林业大学信息科学技术学院,南京210037 [2]江苏开放大学信息工程学院,南京210017

出　　处：《计算机与数字工程》2022年第2期419-423,430,共6页Computer & Digital Engineering

基　　金：国家自然科学基金青年科学基金项目(编号:62102184);江苏省自然科学基金青年科学基金项目(编号:BK20200784);中国博士后科学基金面上项目(编号:2019M661852);江苏省高校自然科学研究面上项目(编号:19KJB520010);江苏省高等学校大学生创新创业训练计划项目“基于高分辨率遥感图像的车流量统计研究”(编号:202010298124Y)资助。

摘　　要：道路是现代交通的主要组成部分,对于管理和更新地理信息系统数据库中的道路信息非常重要。目前,自动提取道路网络的主要数据源为遥感图像数据,但随着近年来遥感影像的地面分辨率不断提高,图像中地物信息愈加丰富,对图像中道路信息的提取难度也随之增大。文章主要展开一种利用机器学习对高分辨率遥感图像的道路提取研究。首先对高分辨遥感图进行预处理,然后对图像进行特征提取,利用BP神经网络对特征进行训练,最后将需要道路提取的高分辨率遥感图区域分割。对每一个区域进行目标检测时,去除图像中非道路区域,并利用形态学方法提取出区域中的道路信息。实验表明,该方法在应对建筑物、植被等对道路提取有干扰时,识别效率明显提升。The road is the main part of modern transportation,which is very important for managing and updating the road information in the global information system database. At present,the main data source of automatic extraction of road networks is in remote sensing image data. However,with the continuous improvement of ground resolution of remote sensing images in recent years,the ground object information in the image is more and more abundant,and the difficulty of extracting road information in the image is increasing. This paper mainly focuses on the research of road extraction from high-resolution remote sensing images using machine learning technique. Firstly,the high-resolution remote sensing image is preprocessed,then the image features are extracted,and the features are trained by BP neural networks. Finally,the high-resolution remote sensing image that needs road extraction is segmented. When each region is detected,the non-road region in the image is removed,and the road information in the region is extracted by this morphological method. Experiments show that the efficiency of this method is significantly improved when dealing with the interference of buildings and vegetation on road extraction.

关 键 词：遥感技术 道路提取 机器学习 神经网络 图像处理 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]