RSI-CB: A Large-Scale Remote Sensing Image Classification Benchmark Using Crowdsourced Data.

Haifeng Li,Jie Chen,Jian Peng,Min Deng,Zhixiang Wu,Chao Tao,Ling Zhao,Xin Dou

doi:10.3390/s20061594

Abstract

Image classification is a fundamental task in remote sensing image processing. In recent years, deep convolutional neural networks (DCNNs) have experienced significant breakthroughs in natural image recognition. The remote sensing field, however, is still lacking a large-scale benchmark similar to ImageNet. In this paper, we propose a remote sensing image classification benchmark (RSI-CB) based on massive, scalable, and diverse crowdsourced data. Using crowdsourced data, such as Open Street Map (OSM) data, ground objects in remote sensing images can be annotated effectively using points of interest, vector data from OSM, or other crowdsourced data. These annotated images can, then, be used in remote sensing image classification tasks. Based on this method, we construct a worldwide large-scale benchmark for remote sensing image classification. This benchmark has large-scale geographical distribution and large total image number. It contains six categories with 35 sub-classes of more than 24,000 images of size pixels. This classification system of ground objects is defined according to the national standard of land-use classification in China and is inspired by the hierarchy mechanism of ImageNet. Finally, we conduct numerous experiments to compare RSI-CB with the SAT-4, SAT-6, and UC-Merced data sets. The experiments show that RSI-CB is more suitable as a benchmark for remote sensing image classification tasks than other benchmarks in the big data era and has many potential applications.

Highlights

Image classification is a fundamental task in remote sensing image processing for which much meaningful research has been carried out [1,2,3,4,5,6]
The experimental results are divided into three parts: classification results based on handcrafted features and Deep convolutional neural networks (DCNNs) and model transfer ability test results
We used the UC-Merced and remote sensing image classification benchmark (RSI-CB) benchmarks with a uniform size of 256 × 256 pixels as experimental data to ensure the fairness of results

Summary

Introduction

Image classification is a fundamental task in remote sensing image processing for which much meaningful research has been carried out [1,2,3,4,5,6]. Deep convolutional neural networks (DCNNs) have been considered as a breakthrough technology since AlexNet [7] achieved impressive results in the ImageNet Challenge [8] in 2012. DCNN on natural image classification, remote sensing experts have introduced DCNN to remote sensing image classification and other recognition tasks [1,28,29,30,31,32], which is the current frontier and pinnacle of remote sensing image processing. The key factors of the DCNN model’s success are its universal approximation ability, large-scale databases (such as ImageNet), and supercomputing ability when powered by graphics processing

Methods

Results

Conclusion