skip to main content


Title: CloudNet: A Deep Learning Approach for Mitigating Occlusions in Landsat-8 Imagery using Data Coalescence
Multi-spectral satellite images that remotely sense the Earth's surface at regular intervals are often contaminated due to occlusion by clouds. Remote sensing imagery captured via satellites, drones, and aircraft has successfully influenced a wide range of fields such as monitoring vegetation health, tracking droughts, and weather forecasting, among others. Researchers studying the Earth's surface are often hindered while gathering reliable observations due to contaminated reflectance values that are sensitive to thin, thick, and cirrus clouds, as well as their shadows. In this study, we propose a deep learning network architecture, CloudNet, to alleviate cloud-occluded remote sensing imagery captured by Landsat-8 satellite for both visible and non-visible spectral bands. We propose a deep neural network model trained on a distributed storage cluster that leverages historical trends within Landsat-8 imagery while complementing this analysis with high-resolution Sentinel-2 imagery. Our empirical benchmarks profile the efficiency of the CloudNet model with a range of cloud-occluded pixels in the input image. We further compare our CloudNet's performance with state-of-the-art deep learning approaches such as SpAGAN and Resnet. We propose a novel method, dynamic hierarchical transfer learning, to reduce computational resource requirements while training the model to achieve the desired accuracy. Our model regenerates features of cloudy images with a high PSNR accuracy of 34.28 dB.  more » « less
Award ID(s):
1931363
NSF-PAR ID:
10448871
Author(s) / Creator(s):
; ; ; ;
Date Published:
Journal Name:
2022 IEEE 18th International Conference on e-Science (e-Science)
Page Range / eLocation ID:
117 to 127
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Cloud detection is an inextricable pre-processing step in remote sensing image analysis workflows. Most of the traditional rule-based and machine-learning-based algorithms utilize low-level features of the clouds and classify individual cloud pixels based on their spectral signatures. Cloud detection using such approaches can be challenging due to a multitude of factors including harsh lighting conditions, the presence of thin clouds, the context of surrounding pixels, and complex spatial patterns. In recent studies, deep convolutional neural networks (CNNs) have shown outstanding results in the computer vision domain. These methods are practiced for better capturing the texture, shape as well as context of images. In this study, we propose a deep learning CNN approach to detect cloud pixels from medium-resolution satellite imagery. The proposed CNN accounts for both the low-level features, such as color and texture information as well as high-level features extracted from successive convolutions of the input image. We prepared a cloud-pixel dataset of approximately 7273 randomly sampled 320 by 320 pixels image patches taken from a total of 121 Landsat-8 (30m) and Sentinel-2 (20m) image scenes. These satellite images come with cloud masks. From the available data channels, only blue, green, red, and NIR bands are fed into the model. The CNN model was trained on 5300 image patches and validated on 1973 independent image patches. As the final output from our model, we extract a binary mask of cloud pixels and non-cloud pixels. The results are benchmarked against established cloud detection methods using standard accuracy metrics.

     
    more » « less
  2. State-of-the-art deep learning technology has been successfully applied to relatively small selected areas of very high spatial resolution (0.15 and 0.25 m) optical aerial imagery acquired by a fixed-wing aircraft to automatically characterize ice-wedge polygons (IWPs) in the Arctic tundra. However, any mapping of IWPs at regional to continental scales requires images acquired on different sensor platforms (particularly satellite) and a refined understanding of the performance stability of the method across sensor platforms through reliable evaluation assessments. In this study, we examined the transferability of a deep learning Mask Region-Based Convolutional Neural Network (R-CNN) model for mapping IWPs in satellite remote sensing imagery (~0.5 m) covering 272 km2 and unmanned aerial vehicle (UAV) (0.02 m) imagery covering 0.32 km2. Multi-spectral images were obtained from the WorldView-2 satellite sensor and pan-sharpened to ~0.5 m, and a 20 mp CMOS sensor camera onboard a UAV, respectively. The training dataset included 25,489 and 6022 manually delineated IWPs from satellite and fixed-wing aircraft aerial imagery near the Arctic Coastal Plain, northern Alaska. Quantitative assessments showed that individual IWPs were correctly detected at up to 72% and 70%, and delineated at up to 73% and 68% F1 score accuracy levels for satellite and UAV images, respectively. Expert-based qualitative assessments showed that IWPs were correctly detected at good (40–60%) and excellent (80–100%) accuracy levels for satellite and UAV images, respectively, and delineated at excellent (80–100%) level for both images. We found that (1) regardless of spatial resolution and spectral bands, the deep learning Mask R-CNN model effectively mapped IWPs in both remote sensing satellite and UAV images; (2) the model achieved a better accuracy in detection with finer image resolution, such as UAV imagery, yet a better accuracy in delineation with coarser image resolution, such as satellite imagery; (3) increasing the number of training data with different resolutions between the training and actual application imagery does not necessarily result in better performance of the Mask R-CNN in IWPs mapping; (4) and overall, the model underestimates the total number of IWPs particularly in terms of disjoint/incomplete IWPs. 
    more » « less
  3. Abstract

    Due to climate change and rapid urbanization, Urban Heat Island (UHI), featuring significantly higher temperature in metropolitan areas than surrounding areas, has caused negative impacts on urban communities. Temporal granularity is often limited in UHI studies based on satellite remote sensing data that typically has multi-day frequency coverage of a particular urban area. This low temporal frequency has restricted the development of models for predicting UHI. To resolve this limitation, this study has developed a cyber-based geographic information science and systems (cyberGIS) framework encompassing multiple machine learning models for predicting UHI with high-frequency urban sensor network data combined with remote sensing data focused on Chicago, Illinois, from 2018 to 2020. Enabled by rapid advances in urban sensor network technologies and high-performance computing, this framework is designed to predict UHI in Chicago with fine spatiotemporal granularity based on environmental data collected with the Array of Things (AoT) urban sensor network and Landsat-8 remote sensing imagery. Our computational experiments revealed that a random forest regression (RFR) model outperforms other models with the prediction accuracy of 0.45 degree Celsius in 2020 and 0.8 degree Celsius in 2018 and 2019 with mean absolute error as the evaluation metric. Humidity, distance to geographic center, and PM2.5concentration are identified as important factors contributing to the model performance. Furthermore, we estimate UHI in Chicago with 10-min temporal frequency and 1-km spatial resolution on the hottest day in 2018. It is demonstrated that the RFR model can accurately predict UHI at fine spatiotemporal scales with high-frequency urban sensor network data integrated with satellite remote sensing data.

     
    more » « less
  4. null (Ed.)
    Mapping invasive vegetation species in arid regions is a critical task for managing water resources and understanding threats to ecosystem services. Traditional remote sensing platforms, such as Landsat and MODIS, are ill-suited for distinguishing native and non-native vegetation species in arid regions due to their large pixels compared to plant sizes. Unmanned aircraft systems, or UAS, offer the potential to capture the high spatial resolution imagery needed to differentiate species. However, in order to extract the most benefits from these platforms, there is a need to develop more efficient and effective workflows. This paper presents an integrated spectral–structural workflow for classifying invasive vegetation species in the Lower Salt River region of Arizona, which has been the site of fires and flooding, leading to a proliferation of invasive vegetation species. Visible (RGB) and multispectral images were captured and processed following a typical structure from motion workflow, and the derived datasets were used as inputs in two machine learning classifications—one incorporating only spectral information and one utilizing both spectral data and structural layers (e.g., digital terrain model (DTM) and canopy height model (CHM)). Results show that including structural layers in the classification improved overall accuracy from 80% to 93% compared to the spectral-only model. The most important features for classification were the CHM and DTM, with the blue band and two spectral indices (normalized difference water index (NDWI) and normalized difference salinity index (NDSI)) contributing important spectral information to both models. 
    more » « less
  5. null (Ed.)
    Abstract. The empirical attribution of hydrologic change presents a unique data availability challenge in terms of establishing baseline prior conditions, as one cannot go back in time to retrospectively collect the necessary data. Although global remote sensing data can alleviate this challenge, most satellite missions are too recent to capture changes that happened long ago enough to provide sufficient observations for adequate statistical inference. In that context, the 4 decades of continuous global high-resolution monitoring enabled by the Landsat missions are an unrivaled source of information. However, constructing a time series of land cover observation across Landsat missions remains a significant challenge because cloud masking and inconsistent image quality complicate the automatized interpretation of optical imagery. Focusing on the monitoring of lake water extent, we present an automatized gap-filling approach to infer the class (wet or dry) of pixels masked by clouds or sensing errors. The classification outcome of unmasked pixels is compiled across images taken on different dates to estimate the inundation frequency of each pixel, based on the assumption that different pixels are masked at different times. The inundation frequency is then used to infer the inundation status of masked pixels on individual images through supervised classification. Applied to a variety of global lakes with substantial long term or seasonal fluctuations, the approach successfully captured water extent variations obtained from in situ gauges (where applicable), or from other Landsat missions during overlapping time periods. Although sensitive to classification errors in the input imagery, the gap-filling algorithm is straightforward to implement on Google's Earth Engine platform and stands as a scalable approach to reliably monitor, and ultimately attribute, historical changes in water bodies. 
    more » « less