Narrow Road Extraction from Remote Sensing Images Based on Super-Resolution Convolutional Neural Network

Abstract

In remote sensing images, it is usually hard to extract narrow roads with only several pixels width. To address this problem, the original remote sensing images are processed with super-resolution to enlarge the details of the narrow roads by a convolutional neural network method. Then the One-Class Support Vector Machine (OCSVM) classifier is applied after super-resolution for exact extraction of narrow roads. Experiments are conducted on an open dataset of remote sensing images to verify the performance of the new method and the results are compared with the method without image super-resolution. The experimental results demonstrate the validity and superiority of the new method.