Advanced remote sensing information system program at ETH Zurich and DLR/DFD Oberpfaffenhofen

Conventional satellite ground segments comprise mission operations and data acquisition systems, data ingestion interfaces, processing capabilities, a catalogue of available data, a data archive, and interfaces for queries and data dissemination. The transfer of data mostly relies on common computer networks or high capacity tapes. Future ground segments have to be capable of handling multi-mission data delivered by Synthetic Aperture Radar (SAR) and optical sensors and have to provide easy user access to support the selection of specific data sets, fuse data, visualize products and to compress data for transmission via Internet. In particular, the search for data sets has to support individual queries by data content and detailed application area (“data mining”) as well as capabilities for automated extraction of relevant features and the application oriented representation of results. In the case of SAR image data, we face an enormous volume of raw data combined with very specific analysis requests posed by the users. In order to reconcile these conflicting aspects we suggest the development of a tool for interactive generation of high level products. This tool shall extract, evaluate, and visualize the significant information from multidimensional image data of geographically localized areas. It shall rely on advanced image compression methods. Thus, the tool will support the monitoring of the Earth’s surface from various perspectives like vegetation, ice and snow, or ocean features. This exchange of information can be understood as image communication, i.e. the transfer of information via visual data. We propose an advanced remote sensing ground segment architecture designed for easier and interactive decision-making applications and a broader dissemination of remote sensing data. This article presents the concept developed at the German Aerospace Center, DLR Oberpfaffenhofen in collaboration with the Swiss Federal Institute of Technology, ETH Zurich, for information retrieval from remote sensing (RS) data. The research line has as very pragmatic goals the design of future RS ground segment systems permitting fast distribution and easy accessibility to the data, real and near real-time applications, and to promote the implementation of data distribution systems. However, the scope of the project is much larger, addressing basic problems of image and information representation, with a large potential of application in other fields: medical sciences, multimedia, interactive television, etc.

Advances in hardware capabilities and big data technologies over the past decade have enabled the application of deep learning techniques to address the challenge of super-resolution in remote-sensing images. Recently, while deep-learning-based methods have outperformed traditional methods, the abundance of information in remote-sensing images creates an imbalance between performance and computational resource consumption in current deep-learning-based methods. This paper introduces a stable super-resolution algorithm based on parameter reconstructive optimization to address these issues. First, based on the stable super-resolution generative adversarial network (SSRGAN), the algorithm employs a generator with networks using residual connections to reconstruct images with enhanced resolution. Next, it extracts content, adversarial, and regularization losses using the discriminator from a stable super-resolution generative adversarial network, which in turn guides the training of the network. Finally, upon completion of training, structural re-parameterization is conducted to optimize the multi-branch trained network into a plain inference network. This inference network can serve as an individual new generator. The results of qualitative and quantitative experimental comparisons with the models Bicubic, super-resolution convolutional neural network, very deep super-resolution convolutional network, deep recursive residual network, super-resolution generative adversarial network (SRGAN), enhanced SRGAN, and stable SRGAN on the Gaofen-5 AHSI satellite dataset suggest that this algorithm achieves improved evaluation indices with a 4× magnification ratio, reaching a peak signal-to-noise ratio of 30.7207 dB and structural similarity index measure of 0.8178. Compared with the trained but unconverted generator, which can also work independently, implementing re-parameterization results in approximately a 10% reduction in the number of parameters, indicating lower resource consumption, while the reconstruction effect is minimally influenced. Furthermore, the super-resolution results exhibit richer detail, increased contrast, and better scene adaptability.

This paper introduces the theories and methods for image fusion in remote sensing. Based on the IHS transform and wavelet packet transform, a scheme of remote-sensing images fusion aiming at protecting image's spectral characteristics is made in this paper. This scheme makes best use of the information in remote-sensing images to be fused and prevents the loss of image information. Through the experiment using SAR image and TM image, all the spectral characteristics, the textural feature and the spatial quality of image have been improved.

The theoretical backscattering model from Fung et al. (1992) was used to simulate the reflection of an electromagnetic wave from a synthetic aperture radar (SAR) on a bare soil surface depending on the soil properties. These data were applied to a neural network (NN) as a training set to retrieve the soil moisture from the backscattering coefficient. Several network configurations were tested and three are discussed. In the next step the trained networks were fed with remotely sensed data from the SAR on board of ERS-2 (European Remote Sensing Satellite) and data from the airborne 'Experimental SAR' (ESAR) built by the Institute of Radiofrequencies of the 'German Aerospace Center' (DLR). The root mean square error (RMSE) between soil moisture calculated from the network and measured in situ during the overflights was below 5% volumetric soil moisture.

To overcome the difficulty of automating and intelligently classifying the ground features in remote-sensing hyperspectral images, machine learning methods are gradually introduced into the process of remote-sensing imaging. First, the PaviaU, Botswana, and Cuprite hyperspectral datasets are selected as research subjects in this study, and the objective is to process remote-sensing hyperspectral images via machine learning to realize the automatic and intelligent classification of features. Then, the basic principles of the support vector machine (SVM) and extreme learning machine (ELM) classification algorithms are introduced, and they are applied to the datasets. Next, by adjusting the parameter estimates using a restricted Boltzmann machine (RBM), a new terrain classification model of hyperspectral images that is based on a deep belief network (DBN) is constructed. Next, the SVM, ELM, and DBN classification algorithms for hyperspectral image terrain classification are analysed and compared in terms of accuracy and consistency. The results demonstrate that the average detection accuracies of ELM on the three datasets are 89.54%, 96.14%, and 96.28%, and the Kappa coefficient values are 0.832, 0.963, and 0.924; the average detection accuracies of SVM are 88.90%, 92.11%, and 91.68%, and the Kappa coefficient values are 0.768, 0.913, and 0.944; the average detection accuracies of the DBN classification model are 92.36%, 97.31%, and 98.84%, and the Kappa coefficient values are 0.883, 0.944, and 0.972. The results also demonstrate that the classification accuracy of the DBN algorithm exceeds those of the previous two methods because it fully utilizes the spatial and spectral information of hyperspectral remote-sensing images. In summary, the DBN algorithm that is proposed in this study has high application value in object classification for remote-sensing hyperspectral images.

Click to increase image sizeClick to decrease image size Additional informationNotes on contributorsHans WalserHans Walser (hwalser@bluewin.ch) received his Ph.D. from the ETH (Swiss Federal Institute of Technology) Zurich in 1975. He is a teacher at the State College of Thurgau and lecturer at the ETH Zurich and the University of Basel. His primary research interests are in geometry. He is also engaged in projects to improve mathematical teaching in Germany.

Remote sensing of Antarctic sea ice is required to characterize properties of the vast sea ice cover to understand its long-term increase in contrast to the decrease of Arctic sea ice. For this objective, the OIB/TanDEM-X Coordinated Science Campaign (OTASC) was successfully conducted in 2017 to obtain contemporaneous and collocated remote sensing data from NASA's Operation IceBridge (OIB) and the German Aerospace Center (DLR) TanDEM-X Synthetic Aperture Radar (SAR) system at X band together with Sentinel-1 and RADARSAT-2 SARs at C band in conjunction with WorldView satellite spectral sensors, surface measurements, and field observations. The Weddell Sea and the Ross Sea were two primary regions while SAR data were also collected over six other regions in the Southern Ocean. Satellite SAR data included both polarimetric and interferometric capabilities to infer snow and sea ice information in three dimensions (3D), while OIB/P-3 aircraft data include snow radar together with altimeter data for snow and sea ice observations in 3D over the Weddell Sea. Across the Ross Sea, IcePOD and AntNZ/York-University flights were carried out together with satellite SAR data acquisitions.

Remote sensing images have been widely used in many scenarios. However, it is difficult to obtain high-resolution images from remote sensing images with the limitations of the sensors. The Zero-Shot Super-Resolution (ZSSR) method constructs the training set only by the input image itself, improving the performance while the blur kernels are unknown. The improvement is limited due to the limited information of the input image itself. In this paper, an enhanced zero-shot super-resolution is proposed to solve the problem of super-resolution(SR) reconstruction when the fuzzy kernel is unknown by downsampling the enhanced images of Content Adaptive Resampler(CAR) network to build the training set, and introducing the Convolutional Block Attention Module (CBAM) and the residual module to expand the network scale and optimizing the network structure. Experimental results on UCMerced_LandUses and other datasets show that, compared with ZSSR, the proposed method has better super-resolution reconstruction results in terms of image detail and image quality.

Satellite Remote Sensing (RS) can be effectively used for the fisheries applications, including stock assessment, management of marine resources and fish harvesting. Further, there is an extensive range of fisheries related applications of RS data, including identifying marine managed areas, detection of harmful algal blooms, bycatch reduction, long term monitoring of fisheries related habitat changes and control of illegal fishing. Applications on RS related to marine fisheries sector in Sri Lanka, seem to be meager. Limited studies have been conducted on the fisheries related ecosystem damage assessment using RS after December 2004 Tsunami disaster. To fulfill the gaps of research on coastal fisheries identified, the author conducted some research on habitat changes such as monitoring of land use changes and water quality at economically important estuarine environments around Sri Lanka using multi temporal satellite imageries. In these studies, several environmental issues have been addressed such as localized eutrophication, disturbance of water circulations and sudden fish kills etc. Furthermore, since satellite RS is a useful tool that could be applied for planning marine fisheries development in Sri Lanka, the National Aquatic Resources Research and Development Agency (NARA) is currently conducting forecasting of fishing ground using satellite data for off-shore/ high seas fishery, providing accurate information on large pelagic fishing grounds considering sea surface temperature and chlorophyll distribution. As illegal, Unreported and Unregulated (IUU) fishing is a serious threat to the long-term economic and environmental sustainability of Sri Lankan oceans, RS techniques exhibit great potential for supporting to develop fisheries monitoring system around Sri Lankan waters which is timely important for control IUU fishing. VIIRS Boat Detection system of NOAA and Global Fishing Watch were discussed as the possible systems that can support the future monitoring and management of Sri Lankan marine fisheries sector. This paper summarizes the present RS related applications to the marine fishery sector of Sri Lanka. Further, it also provides guidelines for sustainable management of the sector that can result from RS, and identify existing applications which can effectively support the marine fishery management of the country.

The German Satellite Data Archive (D-SDA) at the German Aerospace Center (DLR) has been managing largevolume Earth-observation (EO) data in the context of EOmission payload ground segments (PGSs) for more than two decades. Hardware, data management, processing, user access, long-term preservation, and data exploitation expertise are under one roof and interact closely. Upcoming EO-mission PGSs benefit as much from the comprehensive expertise, close interaction, and integrated infrastructure as do in-house scientific application projects requiring access, processing, and archiving of large-volume EO data. Using a number of examples, we will demonstrate how EO data life cycles benefit from the proximity of data management and application scientists and from the extensive operational experience gathered over time.

3D-Particle Tracking Velocimetry (PTV) is one of the most flexible techniques for flow measurement, which allows the determination of three-dimensional velocity fields. Research activities in this field performed by the Institute of Geodesy and Photogrammetry at ETH Zurich for more than a decade have reached a status of an operational and reliable measurement method used in hydrodynamics and space applications. The method is based on the visualization of a flow with small, neutrally buoyant particles and recording of particle image sequences with 3-4 CCD cameras. In cooperation with the Institute of Hydromechanics and Water Resources Management at ETH Zurich further progress has been achieved in the improvement of the existing hardand software solutions. Regarding the algorithmic aspect of the method a new spatiotemporal matching method was developed, implemented and tested on different data sets. In former implementations the determination of the 3D particle positions was separated from the tracking of each particle, while the new method uses a combination of image and object space based information to establish spatio-temporal correspondences between particle positions of consecutive time steps. A system based on 4 CCD cameras is capable of tracking up to 1000 particles at video frequency (25Hz) with a relative accuracy of the velocity vectors of approximately 1:4000 of the field of view. The latest developments of the algorithmic aspects of 3D PTV are described and some examples of the successful application of the method are given in this paper. INTRODUCTION The 3D PTV is a technique for the determination of 3D velocity fields in flows. The existing 3D PTV solution developed at the Institute of Geodesy and Photogrammetry applying a object space based tracking algorithm should be improved in a way that the redundant information in image and object space is exploited more efficiently. The use of image and object space based information in combination with a prediction of the particle motion was thought to lead to enhanced results in the velocity field determination. The most important result to be expected from this work is a substantial increase of the tracking rate in 3D PTV. This is of importance mainly in the context of a Lagrangian analysis of particle trajectories, which can be considered the actual domain of the technique. Long trajectories are an absolute requisite for a Lagrangian flow analysis, as integral time and length scales can only be determined if long correlation lengths have been recorded. In addition, the number of simultaneous trajectories should be large enough to form a sufficient basis for a statistical analysis. Due to interruptions of particle trajectories caused by unsolved ambiguities the number of long trajectories decreases exponentially with the trajectory length. Very long trajectories over hundred and more time instances can so far only be determined if the probability of ambiguities is reduced by a low seeding density, thus concurrently reducing the spatial resolution of the system and the basis for a statistical analysis. A reduction of the trajectory interruptions due to unsolved ambiguities can multiply the yield of long trajectories and thus the usefulness of the results of 3D PTV, which further enlarges the application potential of the technique. Within the framework of a research project of the Swiss National Science Foundation a new spatio-temporal matching algorithm was developed and implemented. The technique has reached a status of an operational and reliable measurement tool used in hydrodynamics and space applications (Becker et al, 1995, Maas et al, 1997, Willneff and Maas, 2000). NOMENCLATURE XO,YO, ZO: 3D coordinates of projective center O Xi,Yi, Zi: 3D coordinates of object point Pi xi, yi: Image coordinates of Pi c: Principle distance of the camera xh, yh: Coordinates of the principle point ω, φ, κ: Rotation angles of the direction of the optical axis Drot: Rotation matrix as function of the rotation angles ω, φ, κ aij: Elements of 3x3 rotation matrix ti: Time step of image sequence

The Earth Observation database is almost on the scale of Zettabyte (1021 Bytes). Produced at a rapid rate, those data also present great diversity, due to the range of sensor types. In such a manner, this kind of data is also classified as Big Data, and present opportunities, such as the possibility to analyze and integrate different data, as well as challenges, mainly regarding storing and processing steps in order to be available to users. The distribution of this database is normally through catalogues, which searching criteria are limited to traditional metadata, as acquisition date, sensor characteristics and geographical localization. Thus, there is a demand for a tool that enables users to search for images based on phenomena in lieu of date or location in a data fusion perspective. In this manner, this work resulted in a Remote Sensing Image Information Mining (ReSIIM) prototype able to make smart searches in big databases based on well-known and basic targets found in Remote Sensing imagery: cloud, cloud shadow, clear land (land area), water, forest, bare soil, built-up and burned area. For that, the aforementioned targets metadata are extract and stored in databases, enabling to refine and boost searches. Besides the spatialization and discussion of the aforementioned targets along Brazil, ReSIIM was assessed according to its accuracy to identify adequate images in databases for Linear Spectral Mixture Model (LSMM) application, which is one of the main methods available to detect burned areas in the Brazilian Amazon. With the adequate search criteria, ReSIIM was able to retrieve all adequate images, indicating its high potential for this and further applications.

With the advancement of remote sensing technology, high-resolution images are widely used in the field of computer vision. However, image quality is often degraded due to hardware limitations and environmental interference. This paper proposes a Residual Separable Super-Resolution Reconstruction Generative Adversarial Network (RSSRGAN) for remote sensing image super-resolution. The model aims to enhance the resolution and edge information of low-resolution images without hardware improvements. The main contributions include (1) designing an optimized generator network by improving the residual dense network and introducing depthwise separable convolutions to remove BN layers, thereby increasing training efficiency—two PatchGAN discriminators are designed to enhance multi-scale detail capture—and (2) introducing content loss and joint perceptual loss on top of adversarial loss to improve global feature representation. Experimental results show that compared to the widely used SRGAN model in remote sensing (exemplified by the satellite-specific SRGAN in this study), this model improves PSNR by approximately 18.8%, SSIM by 8.0%, reduces MSE by 3.6%, and enhances the PI metric by 13.6%. It effectively enhances object information, color, and brightness in images, making it more suitable for remote sensing image super-resolution.

The 2nd International Conference on Computers, Remote Sensing and Aerospace (CRSA 2022) was successfully held in Nagoya, Japan, on August 26-28, 2022. Due to the impact of COVID-19, it was changed to be held online. The conference mainly focused on the latest researches on Computer, Remote Sensing and Aerospace, gathering experts, scholars, and researchers in these fields from all over the world to share research achievements, explore hot issues, and exchange experience and new ideas.In the conference, we were greatly honored to have both Prof. Yanguo Jing from Coventry University, UK and Prof. Jiang Zhu from Tokyo Institute of Technology, Japan to serve as our Conference Chairs. 100 scholars around the world took part in the conference. Divided into three parts, the conference agenda included keynote speeches, oral presentations and online Q&A discussion. Primarily, the keynote speakers were each allocated 30-45 minutes to address their speeches. Then in the next part, oral presentations, the excellent papers we had selected were presented by their authors one by one.Three distinguished professors were invited to hold keynote speeches during the conference. Among them, Prof. Yanguo Jing from Coventry University, UK performed a thought-provoking speech on Artificial Intelligence Applications in Sports Performance. Although Artificial Intelligence (AI) are being used to create new values and opportunities in business and society, there are many challenges in developing these AI applications. In the speech, Yanguo shared his own journey of use artificial intelligence and machine learning to monitor sports performance, and evaluated their effectiveness in assessing energy expenditure during sport exercise. What’s more, Assoc. Prof. Thippa Reddy Gadekallu from Vellore Institute of Technology, India addressed a speech: Federated Learning for Healthcare. He introduced the applications of FL for healthcare informatics, the need for FL in healthcare domain, the fundamentals of FL, and the major motivations behind FL for healthcare applications, etc. And some lessons learnt, challenges, as well as future directions were highlighted. Their insightful speeches had triggered heated discussion. And every participant praised this conference for disseminating useful and insightful knowledge.We are glad to share with you that we received lots of submissions from the conference and we selected a bunch of high-quality papers and compiled them into the proceedings after rigorously reviewing them. These papers feature but are not limited to the following topics: Remote Sensing, Global Positioning and Navigation System, Manned/Unmanned Aviation, High Performance Computing, Algorithms, etc. All the papers have been checked through rigorous review and processes to meet the requirements of publication.We would like to acknowledge all of those who supported CRSA 2022 and made it a great success. Particularly, we would like to thank the Journal of Physics: Conference Series, for all the efforts of its colleagues in publishing this paper volume. We sincerely hope that the CRSA 2022 turned out to be a forum for excellent discussions that enable new ideas to come about, promoting collaborative research.Committee of CRSA 2022List of Committee member is available in this Pdf.

ABSTRACTLand-surface water is an important factor influencing the regional environment and climate and is a key factor in the Tibetan Plateau, which is one of the most sensitive regions to global changes. Because of the high elevation, complex topography, and erratic weather of the Tibetan Plateau, direct measurement of the area of every lake is largely unfeasible. Moreover, complex natural geographic conditions increase the difficulty of image processing and information extraction with remote sensing because they enhance the uncertainty of quantitative data retrieved with satellites. Methods based on spectral features do not generate the expected results of lake area over the Tibetan Plateau due to a lack of distinction between water and other land objects, especially snow, vegetation, and low cloud cover. Therefore, a new method to extract lake area from satellite images in the Tibetan Plateau is needed. In this article, an automatic method was proposed to evaluate lake area during the wet season (from 1 September to 31 October) on the Tibetan Plateau with multi-day Advanced Very High Resolution Radiometer (AVHRR) remote-sensing images on board the Meteorological Operational satellite-A (MetOp-A) satellite. The method considers both spectral and textural features of lakes and does not need a cloud mask as an input. In addition, the Mixture Tuned Matched Filtering (MTMF) algorithm was applied to decompose the mixed pixels to better identify lakes and estimate the lake area. Based on daily lake identifications, the wet season’s lake data were composited with the maximum value composition (MVC) method to determine the lake area. A comparison of our work with the manually interpreted results from Landsat Thematic Mapper (TM) images and observational reports demonstrates the accuracy and reliability of our approach. However, certain factors, i.e. the sensor zenith angle of the polar-orbit satellite and the topography, can affect the lake area extracted from the remote-sensing images.