一种自监督学习框架下的高分辨率遥感影像道路中心线自动提取算法

An Automatic Road Centerline Extraction Algorithm from High Resolution Remote Sensing Images Based on Self-supervised Learning Framework

高分辨率遥感影像的道路提取,对于国土规划、车载导航、应急救灾等方面具有重要的意义。本文提出了一种基于自监督学习框架下的高分辨率遥感影像道路中心线自动提取算法。首先,联合光谱与形状特征设计了一种自动获取道路样本的正样本选取方法;然后引入一种正样本分类框架,基于随机森林构建了一种随机森林正样本单分类器,并得到像元隶属于道路的后验概率;以面向对象的角度联合形状特征与后验概率,得到最终的道路网络;最后利用张量投票算法得到道路中心线。通过与监督、非监督道路提取算法相比,有更好的实验结果,且自动进行,无需人工干预,效率高。

Road extraction from high-resolution remote sensing image has a great significance to land planning and vehicle navigation.This paper proposed an method which can automatic extract road centerline in high resolution remote sensing images based on a self-supervised framework.Firstly,a new method is proposed to extract the positive sample of the road sample by combining the spectral and shape features.Then,a one-classifier framework is introduced,and a random forest positive and unlabeled learning classifier is constructed based on the random forest,then get the posterior probability of the road;the shape feature and the posterior probability are combined to form the final road network in an object-oriented way.Finally,the road centerline is obtained based on the tensor voting algorithm.Compared with the supervised and unsupervised road extraction algorithm,this algorithm achieves better experimental results,and the algorithm is automatic and has the advantages of high efficiency.