Use of satellite imagery to categorize vegetation on the French railway network (SNCF Réseau)

Abstract

In order to ensure the safety and regularity of train traffic, as well as the good condition of the railway infrastructures, SNCF Réseau must control the vegetation with efficiency on the network. For this purpose, the French railway infrastructure manager needs to know which kind of vegetation grows on the network. To make the inventory of the vegetation, a processing chain of very high spatial resolution satellite imagery has been developed. The use of a machine learning algorithm for supervised classification and the automation of the processing chain as well as the reuse of pre-trained models allowed to industrialize at a high temporal frequency the inventory of the vegetation along the French railway network.