A Framework Based on Nesting of Convolutional Neural Networks to Classify Secondary Roads in High Resolution Aerial Orthoimages

Remote sensing imagery combined with deep learning strategies is often regarded as anideal solution for interpreting scenes and monitoring infrastructures with remarkable performancelevels. In addition, the road network plays an important part in transportation, and currently one ofthe main related challenges is detecting and monitoring the occurring changes in order to updatethe existent cartography. This task is challenging due to the nature of the object (continuous andoften with no clearly defined borders) and the nature of remotely sensed images (noise, obstructions).In this paper, we propose a novel framework based on convolutional neural networks (CNNs) toclassify secondary roads in high-resolution aerial orthoimages divided in tiles of 256×256 pixels.We will evaluate the framework’s performance on unseen test data and compare the results withthose obtained by other popular CNNs trained from scratch.