A survey of remote sensing image classification based on CNNs

Abstract

With the development of earth observation technologies, the acquired remote sensing images are increasing dramatically, and a new era of big data in remote sensing is coming. How to effectively mine these massive volumes of remote sensing data are new challenges. Deep learning provides a new approach for analyzing these remote sensing data. As one of the deep learning models, convolutional neural networks (CNNs) can directly extract features from massive amounts of imagery data and is good at exploiting semantic features of imagery data. CNNs have achieved remarkable success in computer vision. In recent years, quite a few researchers have studied remote sensing image classification using CNNs, and CNNs can be applied to realize rapid, economical and accurate analysis and feature extraction from remote sensing data. This paper aims to provide a survey of the current state-of-the-art application of CNN-based deep learning in remote sensing image classification. We first briefly introduce the principles and characteristics of CNNs. We then survey developments and structural improvements on CNN models that make CNNs more suitable for remote sensing image classification, available datasets for remote sensing image classification, and data augmentation techniques. Then, three typical CNN application cases in remote sensing image classification: scene classification, object detection and object segmentation are presented. We also discuss the problems and challenges of CNN-based remote sensing image classification, and propose corresponding measures and suggestions. We hope that the survey can facilitate the advancement of remote sensing image classification research and help remote-sensing scientists to tackle classification tasks with the state-of-art deep learning algorithms and techniques.

Predicting road quality using high resolution satellite imagery: A transfer learning approach

“…Recent improvements in the quality and speed of Convolutional Neural Networks (CNNs) has led to several novel applications in many domains including for satellite imagery [22][23][24][25][26][27]. One of the most prominent examples of this has been recent research into the capability of daytime satellite imagery to predict factors traditionally only collected with on-the-ground surveys, including household income and factors related to health outcomes [28][29][30][31].…”

“…Progress on identifying the limits and opportunities of satellite sources has been swift, with the computer vision and remote sensing (RS) communities collaborating to overcome a number of challenges. A wide range of literature has provided insights into effective technical strategies to overcome these differences; [ 22 ] and [ 32 ] provide a broad overview of the technical objectives and innovations that have emerged over the last few years; we further provide our own review in S1 Appendix .…”

“…Furthermore, Xia, Cao, Wang, and Shang added four types of texture information to satellite images and used the resulting data as the input for a CNN to extract roads, vehicles, and vegetation based on the CNN and conditional random field methods [ 47 ]. These studies helped illustrate that greater texture and spectrum information in multisource data can improve the accuracy of extracting road information from RS images [ 22 ].…”

A Meta-Analysis of Convolutional Neural Networks for Remote Sensing Applications

“…CNNbased approaches have benefited from the recent exponential increase in RS technologies that includes various image types (optical, RADAR, temperature and microwave radiometer, altimeter, etc.) with complex characteristics (high dimensionality, multiple scales, and non-stationary) [37].…”

Análise do Uso e Cobertura da Terra Utilizando Imagens Sentinel-2A e Inteligência Artificial

“…Recentemente, com os avanços na ciência da computação e nas tecnologias GIS, os métodos de classificação supervisionados têm sido amplamente usados, uma vez que são mais robustos e precisos do que as abordagens não supervisionadas (Niemeyer et al, 2014;Jamali, 2019). Os métodos supervisionados incluem algoritmos paramétricos, como máxima verossimilhança, distância mínima e classificadores Bayesianos, e algoritmos não paramétricos, como algoritmos de aprendizado de máquina (Phiri & Morgenroth, 2017), que comparativamente aos métodos paramétricos usuais, têm demonstrado precisão e eficiência superior ao trabalhar com dados de maior complexidade e no mapeamento de grandes áreas (Rodriguez-Galiano et al, 2012;Song et al, 2019), tornando-se cada vez mais populares para classificação de dados de sensoriamento remoto (Belgiu & Drăgu, 2016).…”