Convolutional Neural Network for Satellite Imagery

Abstract

Information extracted from aerial photographs has found applications in different areas including urban planning, crop and forest management, disaster relief, and climate modeling. In many cases information extraction is still performed by human experts, making the process slow, costly, and error prone. The goal of this investigation is to develop methods for automatically extracting the locations of objects such as water resource, forest and urban areas from aerial images. We analyze patterns in land using large-scale satellite imagery data which is available worldwide from third-party providers. For training, given the limited availability of standard benchmarks for remotesensing data, we obtain ground truth land use class labels carefully sampled from open-source surveys, in particular the Urban Atlas land classification dataset of 20 land use classes across 300 European cities. The developed algorithms are based on the implementation of a relatively new approach in the field of deep machine learning - a convolutional neural network. We show how deep neural networks implemented on modern GPUs can be used to efficiently learn highly discriminative image features.

Keywords

• deep learning
• computer vision
• neural network