scholarly journals Satellite image inpainting with deep generative adversarial neural networks

2021 ◽  
Vol 10 (1) ◽  
pp. 121
Author(s):  
Mohamed Akram Zaytar ◽  
Chaker El Amrani
Keyword(s):  
Mean Squared Error ◽  
Satellite Image ◽  
Image Inpainting ◽  
Remote Sensing Data ◽  
Neural System ◽  
Generative Adversarial Networks ◽  
Least Absolute Deviations ◽  
Data Pipeline ◽  
Image Pixels ◽  
Processing Errors

This work addresses the problem of recovering lost or damaged satellite image pixels (gaps) caused by sensor processing errors or by natural phenomena like cloud presence. Such errors decrease our ability to monitor regions of interest and significantly increase the average revisit time for all satellites. This paper presents a novel neural system based on conditional deep generative adversarial networks (cGAN) optimized to fill satellite imagery gaps using surrounding pixel values and static high-resolution visual priors. Experimental results show that the proposed system outperforms traditional and neural network baselines. It achieves a normalized least absolute deviations error of (  &  decrease in error compared with the two baselines) and a mean squared error loss of  (  &  decrease in error) over the test set. The model can be deployed within a remote sensing data pipeline to reconstruct missing pixel measurements for near-real-time monitoring and inference purposes, thus empowering policymakers and users to make environmentally informed decisions.

scholarly journals Operational ocean models in the Adriatic Sea: a skill assessment

Ocean Science ◽  
2008 ◽  
Vol 4 (1) ◽  
pp. 61-71 ◽  
Author(s):  
J. Chiggiato ◽  
P. Oddo
Keyword(s):  
General Circulation ◽  
Mean Squared Error ◽  
Three Dimensional ◽  
Circulation Model ◽  
Remote Sensing Data ◽  
Skill Assessment ◽  
Regional Models ◽  
Operational Systems ◽  
Forecasting System ◽  
The Mediterranean

Abstract. In the framework of the Mediterranean Forecasting System (MFS) project, the performance of regional numerical ocean forecasting systems is assessed by means of model-model and model-data comparison. Three different operational systems considered in this study are: the Adriatic REGional Model (AREG); the Adriatic Regional Ocean Modelling System (AdriaROMS) and the Mediterranean Forecasting System General Circulation Model (MFS-GCM). AREG and AdriaROMS are regional implementations (with some dedicated variations) of POM and ROMS, respectively, while MFS-GCM is an OPA based system. The assessment is done through standard scores. In situ and remote sensing data are used to evaluate the system performance. In particular, a set of CTD measurements collected in the whole western Adriatic during January 2006 and one year of satellite derived sea surface temperature measurements (SST) allow to asses a full three-dimensional picture of the operational forecasting systems quality during January 2006 and to draw some preliminary considerations on the temporal fluctuation of scores estimated on surface quantities between summer 2005 and summer 2006. The regional systems share a negative bias in simulated temperature and salinity. Nonetheless, they outperform the MFS-GCM in the shallowest locations. Results on amplitude and phase errors are improved in areas shallower than 50 m, while degraded in deeper locations, where major models deficiencies are related to vertical mixing overestimation. In a basin-wide overview, the two regional models show differences in the local displacement of errors. In addition, in locations where the regional models are mutually correlated, the aggregated mean squared error was found to be smaller, that is a useful outcome of having several operational systems in the same region.

scholarly journals Hybrid Invariant Local Features for Multiple Satellite Image Matching and Registration

2019 ◽  
Vol 9 (2) ◽  
pp. 309-314
Keyword(s):  
Image Matching ◽  
Feature Matching ◽  
Satellite Image ◽  
Remote Sensing Data ◽  
Local Features ◽  
Feature Descriptor ◽  
Matching Method ◽  
Multiple Images ◽  
Feature Sets ◽  
Precise Relationship

Automatic image registration (IR) is very challenging and very important in the field of hyperspectral remote sensing data. Efficient autonomous IR method is needed with high precision, fast, and robust. A key operation of IR is to align the multiple images in single co-ordinate system for extracting and identifying variation between images considered. In this paper, presented a feature descriptor by combining features from both Feature from Accelerated Segment Test (FAST) and Binary Robust Invariant Scalable Key point (BRISK). The proposed hybrid invariant local features (HILF) descriptor extract useful and similar feature sets from reference and source images. The feature matching method allows finding precise relationship or matching among two feature sets. An experimental analysis described the outcome BRISK, FASK and proposed HILF in terms of inliers ratio and repeatability evaluation metrics.

scholarly journals Atmospheric correction of multispectral satellite imagery

2020 ◽  
pp. 4-11
Author(s):  
Leonid Katkovsky
Keyword(s):  
Land Surface ◽  
Satellite Image ◽  
Atmospheric Correction ◽  
Remote Sensing Data ◽  
Sensor Data ◽  
Physical Parameters ◽  
Landsat 8 ◽  
Optical Parameters ◽  
Outgoing Radiation ◽  
Correction Technique

Atmospheric correction is a necessary step in the processing of remote sensing data acquired in the visible and NIR spectral bands.The paper describes the developed atmospheric correction technique for multispectral satellite data with a small number of relatively broad spectral bands (not hyperspectral). The technique is based on the proposed analytical formulae that expressed the spectrum of outgoing radiation at the top of a cloudless atmosphere with rather high accuracy. The technique uses a model of the atmosphere and its optical and physical parameters that are significant from the point of view of radiation transfer, the atmosphere is considered homogeneous within a satellite image. To solve the system of equations containing the measured radiance of the outgoing radiation in the bands of the satellite sensor, the number of which is less than the number of unknowns of the model, it is proposed to use various additional relations, including regression relations between the optical parameters of the atmosphere. For a particular image pixel selected in a special way, unknown atmospheric parameters are found, which are then used to calculate the reflectance for all other pixels.Testing the proposed technique on OLI sensor data of Landsat 8 satellite showed higher accuracy in comparison with the FLAASH and QUAC methods implemented in the well-known ENVI image processing software. The technique is fast and there is using no additional information about the atmosphere or land surface except images under correction.

scholarly journals TopologyGAN: Topology Optimization Using Generative Adversarial Networks Based on Physical Fields Over the Initial Domain

2020 ◽  
Author(s):  
Zhenguo Nie ◽  
Tong Lin ◽  
Haoliang Jiang ◽  
Levent Burak Kara
Keyword(s):  
Topology Optimization ◽  
Boundary Conditions ◽  
Mean Squared Error ◽  
Sparse Matrices ◽  
Hybrid Network ◽  
Generative Adversarial Networks ◽  
Test Problems ◽  
Generative Adversarial Network ◽  
The Mean ◽  
Physical Fields

Abstract In topology optimization using deep learning, load and boundary conditions represented as vectors or sparse matrices often miss the opportunity to encode a rich view of the design problem, leading to less than ideal generalization results. We propose a new data-driven topology optimization model called TopologyGAN that takes advantage of various physical fields computed on the original, unoptimized material domain, as inputs to the generator of a conditional generative adversarial network (cGAN). Compared to a baseline cGAN, TopologyGAN achieves a nearly 3× reduction in the mean squared error and a 2.5× reduction in the mean absolute error on test problems involving previously unseen boundary conditions. Built on several existing network models, we also introduce a hybrid network called U-SE(Squeeze-and-Excitation)-ResNet for the generator that further increases the overall accuracy. We publicly share our full implementation and trained network.

scholarly journals REMOTE SENSING AND GIS FOR LAND COVER CLASSIFICATION AND CHANGE DETECTION IN GAZIPUR SADAR, BANGLADESH DURING 1973-2017

2018 ◽  
Vol 5 (2) ◽  
pp. 215
Author(s):  
Md Arafat Hassan ◽  
Rakibul Islam ◽  
Rehnuma Mahjabin
Keyword(s):  
Remote Sensing ◽  
Supervised Classification ◽  
Agricultural Land ◽  
Satellite Image ◽  
Remote Sensing Data ◽  
Remote Sensing And Gis ◽  
Vegetation Coverage ◽  
Accurate Information ◽  
Sal Forest ◽  
Shed Light

This paper has been developed to capture the land coverage change in Gazipur Sadar Upazila with the help of remote sensing data of 44 years from 1973 to 2017. After acquiring the study area image of 1973, 1991, 2006 and 2017 supervised classification method has been used to get the accurate information from the satellite image and the whole outcome has been transformed into measurable unit (sq km) and graphs. The accuracy of land coverage was ranged from 85% to 89%. The outcome says that the acceleration of economic growth and pressure of huge population took a heavy toll on the vegetation coverage which decreased -199.7%. People are destroying vegetation coverage for building up settlements and infrastructure. In the year 2017, the map shows that the built-up area increased 312.9% for industry, settlement and agricultural purpose. Moreover agricultural land also drops down from 42% to 32%.  The rapid rate of decreasing vegetation coverage and small amount of existing vegetation coverage only 57 sq km (in 2017) is a red alert for the region. The Sal forest and other special flora species of that region is valuable resource for environment. This paper shed light on the fact that it is urgent to protect vegetation coverage so it will help the authority to make good policies and use other techniques to save vegetation coverage.

scholarly journals A Review on Remote Sensing Data Fusion with Generative Adversarial Networks (GAN)

2021 ◽  
Author(s):  
Peng Liu
Keyword(s):  
Remote Sensing ◽  
Data Fusion ◽  
Remote Sensing Data ◽  
Generative Adversarial Networks ◽  
Future Research ◽  
Observation Data ◽  
Sensing Data ◽  
Advantages And Disadvantages ◽  
Adversarial Networks ◽  
Active Research

In the past decades, remote sensing (RS) data fusion has always been an active research community. A large number of algorithms and models have been developed. Generative Adversarial Networks (GAN), as an important branch of deep learning, show promising performances in variety of RS image fusions. This review provides an introduction to GAN for remote sensing data fusion. We briefly review the frequently-used architecture and characteristics of GAN in data fusion and comprehensively discuss how to use GAN to realize fusion for homogeneous RS data, heterogeneous RS data, and RS and ground observation data. We also analyzed some typical applications with GAN-based RS image fusion. This review takes insight into how to make GAN adapt to different types of fusion tasks and summarizes the advantages and disadvantages of GAN-based RS data fusion. Finally, we discuss the promising future research directions and make a prediction on its trends.

scholarly journals Predicting Channel Conveyance and Characterizing Planform Using River Bathymetry via Satellite Image Compilation (RiBaSIC) Algorithm for DEM-Based Hydrodynamic Modeling

2020 ◽  
Vol 12 (17) ◽  
pp. 2799
Author(s):  
Md N M Bhuyian ◽  
Alfred Kalyanapu
Keyword(s):  
Mean Squared Error ◽  
Satellite Image ◽  
Shuttle Radar Topography Mission ◽  
Willamette River ◽  
Bathymetric Data ◽  
Digital Elevation ◽  
Channel Slope ◽  
Dem Accuracy ◽  
Landsat Satellite ◽  
Discharge Estimation

Digital Elevation Models (DEMs) are widely used as a proxy for bathymetric data and several studies have attempted to improve DEM accuracy for hydrodynamic (HD) modeling. Most of these studies attempted to quantitatively improve estimates of channel conveyance (assuming a non-braided morphology) rather than accounting for the actual channel planform. Accurate representation of river conveyance and planform in a DEM is critical to HD modeling and can be achieved with a combination of remote sensing (e.g., satellite image) and field data, such as water surface elevation (WSE). Therefore, the objectives of this study are (i) to develop an algorithm for predicting channel conveyance and characterizing planform via satellite images and in situ WSE and (ii) to estimate discharge using the predicted conveyance via an HD model. The algorithm is named River Bathymetry via Satellite Image Compilation (RiBaSIC) and uses Landsat satellite imagery, Shuttle Radar Topography Mission (SRTM) DEM, Multi-Error-Removed Improved-Terrain (MERIT) DEM, and observed WSE. The algorithm is tested on four study areas along the Willamette River, Kushiyara River, Jamuna River, and Solimoes River. Channel slope and predicted hydraulic radius are subsequently estimated for approximating Manning’s roughness factor. Two-dimensional HD models using DEMs modified by the RiBaSIC algorithm and corresponding Manning’s roughness factors are employed for discharge estimation. The proposed algorithm can represent river planform and conveyance in single-channeled, meandering, wandering, and braided river reaches. Additionally, the HD models estimated discharge within 14–19% relative root mean squared error (RRMSE) in simulation of five years period.

scholarly journals Evaluation of the Multi-Angle Implementation of Atmospheric Correction (MAIAC) Aerosol Algorithm for Himawari-8 Data

2019 ◽  
Vol 11 (23) ◽  
pp. 2771 ◽  
Author(s):  
Lu She ◽  
Hankui Zhang ◽  
Weile Wang ◽  
Yujie Wang ◽  
Yun Shi
Keyword(s):  
Mean Squared Error ◽  
Atmospheric Correction ◽  
Remote Sensing Data ◽  
Japan Meteorological Agency ◽  
Absolute Bias ◽  
Moderate Resolution ◽  
Resolution Imaging ◽  
Moderate Resolution Imaging Spectroradiometer ◽  
Highly Correlated ◽  
New Generation

Himawari-8, operated by the Japan Meteorological Agency (JMA), is a new generation geostationary satellite that provides remote sensing data to retrieve atmospheric aerosol optical depth (AOD) at high spatial (1 km) and high temporal (10 min) resolutions. The Geostationary- National Aeronautics and Space Administration (NASA) Earth exchange (GeoNEX) project recently adapted the multiangle implementation of atmospheric correction (MAIAC) algorithm, originally developed for joint retrieval of AOD and surface anisotropic reflectance with the moderate resolution imaging spectroradiometer (MODIS) data, to generate Earth monitoring products from the latest geostationary satellites including Himawari-8. This study evaluated the GeoNEX Himawari-8 ~1 km MAIAC AOD retrieved over all the aerosol robotic network (AERONET) sites between 6°N–30°N and 91°E–127°E. The corresponding JMA Himawari-8 AOD products were also evaluated for comparison. We only used cloud-free and the best quality satellite AOD retrievals and compiled a total of 16,532 MAIAC-AERONET and 21,737 JMA-AERONET contemporaneous pairs of AOD values for 2017. Statistical analyses showed that both MAIAC and JMA data are highly correlated with AERONET AOD, with the correlation coefficient (R) of ~0.77, and the root mean squared error (RMSE) of ~0.16. The absolute bias of MAIAC AOD (0.02 overestimation) appears smaller than that of the JMA AOD (0.05 underestimation). In comparison with the JMA data, the time series of MAIAC AOD were more consistent with AERONET AOD values and better capture the diurnal variations of the latter. The dependence of MAIAC AOD bias on scattering angles is also discussed.

scholarly journals Image Inpainting Based on Generative Adversarial Networks

IEEE Access ◽  
2020 ◽  
Vol 8 ◽  
pp. 22884-22892 ◽  
Author(s):  
Yi Jiang ◽  
Jiajie Xu ◽  
Baoqing Yang ◽  
Jing Xu ◽  
Junwu Zhu
Keyword(s):  
Image Inpainting ◽  
Generative Adversarial Networks ◽  
Adversarial Networks
Sign in / Sign up

Export Citation Format

Share Document