Clouds Classification from Sentinel-2 Imagery with Deep Residual Learning and Semantic Image Segmentation

Cheng-Chien Liu,Yu-Cheng Zhang,Chien-Chih Lai,Ming-Hsun Ko,Pei-Yin Chen,Yi-Hsin Chen,Ji-Hong Cheng

doi:10.3390/rs11020119

Abstract

Detecting changes in land use and land cover (LULC) from space has long been the main goal of satellite remote sensing (RS), yet the existing and available algorithms for cloud classification are not reliable enough to attain this goal in an automated fashion. Clouds are very strong optical signals that dominate the results of change detection if they are not removed completely from imagery. As various architectures of deep learning (DL) have been proposed and advanced quickly, their potential in perceptual tasks has been widely accepted and successfully applied to many fields. A comprehensive survey of DL in RS has been reviewed, and the RS community has been suggested to be leading researchers in DL. Based on deep residual learning, semantic image segmentation, and the concept of atrous convolution, we propose a new DL architecture, named CloudNet, with an enhanced capability of feature extraction for classifying cloud and haze from Sentinel-2 imagery, with the intention of supporting automatic change detection in LULC. To ensure the quality of the training dataset, scene classification maps of Taiwan processed by Sen2cor were visually examined and edited, resulting in a total of 12,769 sub-images with a standard size of 224 × 224 pixels, cut from the Sen2cor-corrected images and compiled in a trainset. The data augmentation technique enabled CloudNet to have stable cirrus identification capability without extensive training data. Compared to the traditional method and other DL methods, CloudNet had higher accuracy in cloud and haze classification, as well as better performance in cirrus cloud recognition. CloudNet will be incorporated into the Open Access Satellite Image Service to facilitate change detection by using Sentinel-2 imagery on a regular and automatic basis.

Highlights

Detecting changes in land use and land cover (LULC) from space has long been the main goal of satellite remote sensing (RS) [1,2]
This design effectively avoided the loss of spatial information and achieved higher cirrus cloud recognition accuracy compared to existing methods
The results indicate that CloudNet surpassed an fully convolutional neural network (FCN), Deeplab v3+, and an scene classification (SCL) for all indicators: true positive rate (TPR), true negative rate (TNR), precision, IoU, IoU, mIoU, kappa, and pixel accuracy

Summary

Introduction

Detecting changes in land use and land cover (LULC) from space has long been the main goal of satellite remote sensing (RS) [1,2]. High-quality imagery opened from Landsat-8 and Sentinel-2 programs has provided an unprecedented source for satellite observation, various approaches and tools of change detection have been proposed and developed [3,4], and the potential of time series observations has been demonstrated clearly [5,6], near real-time changes in LULC detected from spaceborne observations in a fully automated fashion are still not reliable Clouds and their cast shadows found inevitably in an optical image are both very strong signals that represent a large fraction of changes by comparing two images. From the perspective of developing an automatic change detection system for LULC, the key task is to seek a reliable approach in classifying clouds first

Methods

Results

Conclusion