DLR Portal
Home
|
Sitemap
|
Kontakt
|
Barrierefreiheit
Impressum und Nutzungsbedingungen
Datenschutz
Cookies & Tracking
|
English
Sie sind hier:
Home
:
Aktuelles
:
Wissenschaftliche Kolloquien
:
Institutsseminar 2023
Erweiterte Suche
Aktuelles
Wissenschaftliche Kolloquien
Standortkolloquium
Institutsseminar
Institutsseminar 2024
Institutsseminar 2023
Institutsseminar 2022
Institutsseminar 2021
Institutsseminar 2020
Institutsseminar 2019
Institutsseminar 2018
Institutsseminar 2017
Institutsseminar 2016
Institutsseminar 2015
Institutsseminar 2014
Institutsseminar 2013
Institutsseminar 2012
Institutsseminar 2011
Institutsseminar 2010
Institutsseminar 2009
Institutsseminar 2008
Institutsseminar 2007
Institut
Abteilungen
Compact Test Range
Projekte
TanDEM-X Science
Tandem-L Science
Helmholtz-Allianz
Publikationen des Instituts
Preisverleihungen
Stellenangebote
Angebote zu Praktikum, Bachelor-, Masterarbeiten und Dissertationen
Angebote für Schüler und Studenten
Archiv - Konferenzen
Software
Anreise
Vorträge im Institutsseminar 2023
Lanqing Huang "Sea Ice Topographic Retrieval from TanDEM-X Imagery"
Dienstag, 12. Dezember 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Understanding the height and roughness of polar sea ice is essential. The single-pass interferometric synthetic aperture radar (InSAR) can be used measure the height of drifting sea ice, but this method can be inaccurate because radar signals can penetrate snow and ice. Our research introduces a new method to calculate sea ice height and roughness using InSAR images for different types of ice. We test the accuracy of this method in a 200 by 19 km area and applied it to study sea ice in the Weddell and Ross Seas in Antarctica. We analyze how sea ice heights and roughness vary in these regions and investigate the statistical distributions of ice heights, which differ depending on ice thickness. These findings improve our understanding of sea ice and its impact on the environment and climate in the Weddell and Ross Seas. Additionally, this method can potentially study how sea ice changes over time when applied to time-series data.
Dr. Delwyn Moller and Dr. Brian Pollard "An Overview of Radar Remote Sensing Collaborative Initiatives: Rongowai and Takahé"
Freitag, 01. Dezember 2023 um 10.30 h
Dieser Vortrag findet virtuell statt!
Abstract:
The Rongowai project (te reo Maori for “to sense water”) is a world-first initiative based in New Zealand that aims to advance earth observation using next-generation Global Navigation Satellite Systems Reflectometry (GNSS-R) sensors. A NASA-developed sensor has been mounted on an Air New Zealand Q300 passenger aircraft and is collecting land-surface and coastal data daily between airport hubs across the country. Rongowai's new mission model presents a powerful new paradigm for sensing. It places emphasis on sustainability without compromising quality observations that provide unique insights into the Earth's essential climate variables collecting measurements throughout New Zealand with unprecedented spatio-temporal sampling. We will present some initial results from the first year of operations in addition to ongoing research development, especially in the light of the additional of first-of-a-kind polarimetric capability to the GNSS-R measurements. This new capability is directly relevant to ESA’s HydroGNSS as well in-development satellites from commercial providers Spire and Muon. Proposed plans to support calibration and validation of new geophysical retrievals will also be presented to include utilising targeted airborne Polarimetric SAR imagery to support the GNSS-R analysis. We follow this with an introduction to a dual-platform, interferometric Ka-band mission concept development for advancing maritime security and oceanic and inland water body environmental sensing. The Takahé (Tandem Ka and High-altitude Explorer and also a native New Zealand bird) proposes to use temporal correlation properties of a water surface contrasted with surfaced object or objects to enhance detectability of items of interest, natural or anthropogenic, from that of a vast expanse of water. Once detected further interrogation can be engaged by tasking imaging capability, including radar imaging and Interferometric SAR from long endurance stratospheric platforms. In our talk we will present some of the founding measurements to support our premise of using mm-wave interferometric temporal correlation as a robust discriminator. We will further these observations with predictions and sensitivity analyses for a simulated ocean surface. The strawman mission design, with an emphasis on the technological drivers for the tandem satellite portion of Takahé will be presented with clear synergies with the SkaDI mission in addition to DLR areas of research and development. The Takahē stratospheric segment involves a multi-baseline interferometric synthetic aperture radar (InSAR) for persistent monitoring of regional targets. We conclude this talk with an introduction to the stratospheric InSAR being developed under a NASA Earth Science Technology office program. This X-band InSAR system, based on a fully-steerable active array yet weighing under 7kg, has recently been deployed on a low altitude drone, and is planned for first stratospheric flights in the Spring of 2024. We discuss some of the unique challenges of stratospheric InSAR, including managing the possibly long and complex trajectories of stratospheric vehicles through radar-derived position, orientation, and timing.
David Chaparro Danon "Analysis of water relations in the soil-plant-atmosphere continuum using a multi-satellite approach"
Mittwoch, 29. November 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Passive microwave satellite missions inform on the available soil moisture (SM) for plant water uptake, as well as on the attenuation that the vegetation exerts over the land microwave emissions. The latter is measured by the parameter vegetation optical depth (VOD), which is a function of the vegetation moisture, canopy structure, and biomass. During my 2-year postdoctoral research with DLR I have addressed two main research topics using these and other complementary variables. First, I have developed a new retrieval algorithm which estimates the vegetation moisture content in gravimetric units (kg water / kg fresh biomass). This is done by isolating the water component from the VOD signal. To do so, complementary information is needed to account for the biomass and structure components. This information is based on canopy height measurements from GEDI and on normalized radar backscatter measurements from Sentinel-1. Second, I am studying the time-lagged correlations between microwave-derived SM, VOD, and atmospheric vapor pressure deficit (VPD) to perform a detailed global-scale analysis of the Soil Plant Atmosphere Continuum (SPAC). Both research lines together open a new path towards Earth Observation-based enhanced assessment of soil and plant conditions, drought impacts, and SPAC water pools and fluxes and their influence on climate extremes.
Francisco Jose de Lavor Pereira "RFI Detection in X-Band SAR Imagery using Deep Learning" and Lucas Brianese "A Detailed Analysis of Variable Radar Backscatter Characteristics of the Amazon Rainforest"
Dienstag, 28. November 2023 um 14.00 h
Die Vorträge finden virtuell statt!
RFI Detection in X-Band SAR Imagery using Deep Learning
Abstract:
In this study, we investigate and design an automated technique to detect Radio Frequency Interference (RFI) in X-band Synthetic Aperture Radar (SAR) quicklook images. RFI patterns significantly reduce the quality and utility of SAR images. Therefore, an efficient RFI detection method plays an essential role in the selection of clean SAR images for current and future missions. Using multi-year imagery acquired by TerraSAR-X and TanDEM-X, we introduced an algorithm based on peak amplitude analysis to pre-classify and screen the data. Subsequently, we manually reviewed the pre-classified images to curate datasets with RFI and without RFI. These curated image sets served as the basis for training a Deep Learning Model for image classification, utilizing an architecture based on the Xception Convolutional Neural Network. Even though the limited amount of RFI-labeled data within our dataset poses a major challenge to achieving effective model training it is possible to develop a reliable RFI detection model using novel techniques to overcome this limitation.
A Detailed Analysis of Variable Radar Backscatter Characteristics of the Amazon Rainforest
Abstract:
The Institute’s two twin German Earth observation SAR satellites, TerraSAR-X and TanDEM-X, have been delivering high performance SAR data at X-band for well over a decade while flying in close formation. Prerequisite for achieving this high performance has been the establishment and maintenance of a state-of-the-art SAR System Calibration and long-term system monitoring (LTSM) concept. This relies on the continuous evaluation of standardized data takes. In particular, the verification and long-term system monitoring of the radar front end uses distributed target scenes which appear homogeneous at the radar’s frequency while providing sufficient signal-to-noise-ratio. Throughout the operational mission phase, LTSM acquisitions have been regularly acquired at five different sites throughout the Amazon with routine evaluations showing that the requirements on radiometric stability have been met throughout the mission lifetime. Enabled by a now 15-year long time-series, more detailed analyses reveal temporal and spatial variations even smaller than the originally specified performance limits. After briefly addressing different masking methods for excluding non-forest pixels from the evaluation, this presentation will discuss a detailed quantitative analysis of five unique time-series of more than a decade of data acquisitions over five distributed target sites throughout the Amazon. In particular we show how the mean gamma_0 values vary on different time scales (e.g., long-term trends, seasonality, intra-day variability).
Florian Bischeltsrieder "On the Applicability of the Quantum Advantage: An Engineering Perspective on Quantum Radar"
Donnerstag, 23. November 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Quantum radar is an emerging technology that has attracted increasing interest over the past years. Recent progress has demonstrated that the theoretically predicted advantage over classic radar implementations can be obtained in experiment. In this talk, a brief summary of the currently discussed and experimentally feasible concept of quantum radar is given. We explore the radar task for which quantum technology might offer an improvement and discuss the exact mapping of the real-world operating domain in detail.
Reuben Solomon Katz "ANALYSIS AND SIMULATION OF FORMATION CONTROL STRATEGIES FOR MULTISTATIC SAR CONSTELLATIONS"
Dienstag, 21. November 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Presently, most missions involving synthetic aperture radars (SAR) are characterized by employing substantial monolithic satellite architectures. However, the effectiveness of these large monolithic satellites in enhancing data quality in terms of resolution, coverage, and the visualization of dynamic scenes is limited. A notable recent advancement has been the implementation of satellite constellations, exemplified by the CapellaX-SAR and ICEYE constellations, which have already made significant strides in terms of resolution improvement. Furthermore, the TerraSAR-X and Tandem-X missions, launched in 2010 by the German Aerospace Center (DLR), adopt a formation flying approach, which enables the analysis of dynamic scenes. Despite this advantage, they do not achieve enhanced coverage. Within this framework, significant potential exists for SAR missions through the strategic implementation of formation flying. Numerous spacecraft configurations and the utilization of larger satellite formations are possible to be explored with its set of challenges. These encompass the development of real-time guidance and navigation systems, the adept implementation of control techniques, and the estimation of fuel requisites for both formation flying and station keeping, along with astute launch strategies for multiple satellites. In this thesis, these challenges will be analyzed.
Atila Arantes A. Diniz "Coregistration of Highly Ambiguous SAR Images for Multi-Baseline Distributed SAR Interferometry"
Dienstag, 14. November 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Distributed synthetic aperture radar (SAR) systems are promising solutions to the demand for frequent observation to study and monitor dynamic systems on the Earth. SAR images formed from the data of single receiver in a distributed SAR system are highly ambiguous, which renders conventional coregistration algorithms unusable. This work proposes a coregistration scheme for highly ambiguous SAR images that exploits the difference in the Doppler spectra of the azimuth ambiguities and the main signal to reject the former. The local cross-correlation between the images is evaluated, filtered to a number of Doppler bands, and combined to reject the ambiguous components, revealing the main signal from which the coregistration shifts are finally estimated. The technique was tested using TanDEM-X images with added strong azimuth ambiguities. This novel coregistration scheme enables the use of distributed SAR systems with sub-Nyquist pulse repetition frequency for interferometric applications.
Roman Guliaev "Integrating Polarimetric, Interferometric and Tomographic SAR Information for Improving Forest Structure Estimates"
Donnerstag, 9. November 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
The reconstruction of the vertical reflectivity profile is a central element in the 3D characterization of forests, as well as for the quantitative estimation of key forest parameters such as forest height and biomass. However, the reconstruction and use of vertical reflectivity profiles have so far been performed either in a purely tomographic or interferometric scheme, and almost exclusively in the context of measurements from a single SAR mission or instrument. There is currently no general framework that enables the synergistic reconstruction and/or use of vertical reflectivity profiles through combined InSAR, PolInSAR and TomoSAR measurements. At the same time, an approach for the reconstruction and/or use of vertical reflectivity profiles in the context of multiple missions or frequencies is also lacking. However, the technical evolution of SAR configurations that allow the implementation of multiple modes, and the increasing number of missions that provide measurements sensitive to forest structure at different frequencies, make the development of frameworks that allow the combination of multi-modal, multi-frequency and multi-mission data essential. In this sense, this talk addresses the design and development of frameworks that consider multi-modal and multimission approaches for the reconstruction and/or use of vertical reflectivity profiles in the context of actual mission scenarios.
Philipp Brücker "DRONAR-SAR-Prozessor"
Dienstag, 07. November 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
The rapid developments in Unmanned Aerial Vehicles (UAVs) achieved in recent years, in combination with highly integrated ultra-wideband, lightweight radar systems, allow for innovative synthetic aperture radar concepts in close-range high-resolution imaging applications. Nearly arbitrary flight trajectories are possible using UAVs and in combination with moderate system prices, these advantages lead to new methods in sensing applications with respect to imaging modes, revisiting times and cost efficiency. Besides such capabilities, there are still many open questions for implementing this technology. In the talk, the current status and recent results of the UAV-based SAR sensor DRONAR, being developed in the institute`s microwave sensors group, are presented in detail.
Johannes Rönner "Filtering and Integration Techniques for the Estimation of Ionospheric Signatures with an Autofocus"
Donnerstag, 26. Oktober 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
The ionosphere leaves a signature in low frequency synthetic aperture radar images. In missions Biomass, the spatially fast changing (turbulent part) ionosphere introduces high frequency phase errors that are seen are defocusing. In the framework of the mission, we are working on a Mapdrift autofocus that would correct for phase errors and at the same time recover the turbulent component of the ionosphere. In this work we present some techniques to filter and integrate the observations derived by the Mapdrift and introduce hints to extract geophysical parameters from the defocusing signatures.
Prof. Fabrizio Lombardini "From 3D SAR tomography to "5D" SAR imaging"
Mittwoch, 27. September 2023 um 13.30 h
Dieser Vortrag findet virtuell statt!
Abstract:
After recalling the concept of full 3D SAR imaging by tomographic processing of multibaseline data, in a spectral vision and with a few insights on superresolution in a vector projection interpretation, the seminar touches the basics of a more general and recent multidimensional SAR imaging framework. The concept of 4D (3D+Time) SAR imaging, i.e. “differential tomography”, bridging the gap between differential interferometry and 3D tomography, is illustrated. A 5D information extraction extension is also hinted, and a few urban layover and forest sample results are discussed.
Marcel Stefko (ETH) "Development of a bistatic real-aperture polarimetric radar system and its applications in snow-covered environments"
Dienstag, 26. September 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Bistatic radar is a method which employs a spatially separated transmitter-receiver pair in order to observe backscatter occurring under a particular non-zero scattering angle. Compared to the more common monostatic radar systems (which use a co-located transmitter and receiver), bistatic radar systems (especially those acquiring with bistatic angles larger than 1 degree) are considered to be specialized tools which are more suitable for certain specific purposes, at cost of higher complexity. The special-purpose character and higher complexity of bistatic systems cause a lower availability of such systems, and thus also of bistatic radar datasets. In the Earth Observation domain, such datasets may be needed to, e.g., explore non-reciprocal scattering processes which do not occur in the monostatic regime, or investigate phenomena with specific bistatic signatures, such as opposition effects or specular reflection. KAPRI is a ground-based real-aperture Ku-band polarimetric radar system operated at ETH, based on the Gamma GPRI. It was specially modified to enable using two KAPRI devices in a bistatic configuration, and thus is a useful tool for exploration of the bistatic parameter space, and investigation of above-mentioned bistatic phenomena. Its flexibility allows for a relatively low-cost approach to preliminary investigations of novel bistatic acquisition methods, and its short revisit time (on the order of minutes) allows observation of fast-occurring phenomena which might be otherwise adversely affected by rapid temporal decorrelation. In this talk, the development of KAPRI's bistatic capabilities and subsequent investigations of snow-covered environments with KAPRI will be presented. In the first part, I will present the bistatic operation mode of KAPRI, and a novel polarimetric calibration target VSPARC which was employed to achieve full-polarimetric calibration in the bistatic regime. I will then present findings from a multi-seasonal acquisition campaign at the High-Altitude Research Station Jungfraujoch, where the calibrated system was deployed to acquire a combined monostatic and bistatic full-polarimetric time series dataset of radar observations of the surface of the Great Aletsch Glacier. In the third part, I will present how KAPRI was used to observe the coherent backscatter opposition effect (CBOE) in terrestrial seasonal snow in Davos, Switzerland. Finally, I will elaborate on possible future directions of exploration of bistatic radar applications in snow-covered environments.
Florian Grabs "Azimuth Ambiguity Mitigation in SAR"
Mittwoch, 19. September 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Azimuth ambiguities occur during the acquisition of synthetic aperture radar (SAR) data as alias effects and overlap with the useful signal, thus degrading its quality. In the case of data recorded with the Flugzeug-SAR (F-SAR) system of the German Aerospace Center (DLR), strong, point like scatterers cause the worst azimuth ambiguities. They become visible as stains in the final images which are one product of SAR missions.
A new method to mitigate azimuth ambiguities is proposed in this work. A matched filter for refocusing azimuth ambiguities is designed. In this process, the original point-like objects are restored from the stains. The energy of ambiguities originating from point like scatterers can then be removed with minimal impact on the non-ambiguous data. Therefore a masking operation was optimized.
The algorithm was tested with data of the DLR’s F-SAR system and showed promising results. After the removal of the azimuth ambiguities, structures overlaid by the stains have become visible again. A mitigation of azimuth ambiguities is not only important for a good image quality but can also improve all subsequent tasks like interferometric data processing or classification tasks.
Benjamin Chauvel "Unsupervised learning for forest mapping from spaceborne interferometric SAR imagery"
Dienstag, 22. August 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
The baseline approach for forest mapping with TanDEM-X data is based on a supervised fuzzy clustering algorithm, which uses as main input feature the volume decorrelation factor. This approach was used for the generation of the global TanDEM-X Forest/Non-Forest map at 50 m resolution. However, deep learning methods have already shown first improved results in mapping forests at large scale using TanDEM-X interferometric data.
In this study, we explore the potential of using a U-Net-like architecture for forest mapping at a higher resolution of 6 m with TanDEM-X interferometric data. To address the challenge of limited reference data at this resolution, we investigate a self-supervised learning (SSL) approach. Through a study conducted over temperate forest in Pennsylvania, USA, where a forest map at 1m resolution is available, the results of a fully-supervised approach with a SSL one, have been compared. The impact of various SSL parameters, such as the SSL task and available data volume, is evaluated. The findings from this study will enable us to develop a robust process applicable to areas lacking reliable reference data, particularly in regions like the Amazon rainforest, where accurate forest mapping is vital.
Francesca Scala "Multiple satellites formation flying for Earth observation applications in low Earth orbit"
Donnerstag, 03. August2023 um 14.30 h
Dieser Vortrag findet virtuell statt!
Abstract:
The introduction of constellations and distributed space systems has the potential to improve data quality for Earth observation. In the synthetic aperture radar (SAR) context, the TanDEM-X successfully implements a distributed SAR system. On the other hand, missions carrying passive interferometers, such as the ESA’s Soil Moisture and Ocean Salinity (SMOS) or NASA’s Soil Moisture Active Passive (SMAP), are based on a monolithic satellite architecture. Although both missions are still operational, they have significantly exceeded their lifetime, and future mission concepts are needed. Introducing distributed systems for passive interferometry would adequately address the need for high-resolution passive L-band observations. There are several research and technological challenges connected to such a concept. First, vehicles should be separated by few tens of meters to perform passive interferometry with a swarm/formation, which triggers many difficulties in operating such space systems. Second, precise and autonomous navigation and control techniques are required, even for the real-time relative guidance, navigation, and control (GNC) sub-system. This work starts from these challenges and proposes two future mission concepts capable of achieving high spatial and radiometric resolution as part of a study guided by ESA. These two multi-satellites mission concepts benefit from using relative GNC strategies, including onboard autonomy and robust control. The results of the analyses show the feasibility of the studies and enable the design of future multi-satellite systems for high-resolution passive interferometry.
Andreas Heinzel "Konzepte zur SAR-tomographischen Abbildung vergrabener Objekte"
Donnerstag, 27. Juli 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Die Detektion von explosiven Kampfmitteln stellt heutige Sensorsysteme vor große Herausforderungen. Durch weltweit anhaltende Konflikte steigt die Anzahl der Landminen kontinuierlich an, was insbesondere für die Bevölkerung in Krisengebieten ein erhebliches Risiko darstellt. In diesem Vortrag werden Herausforderungen und Lösungen bei der Detektion von vergrabenen explosiven Kampfmitteln mit Hilfe von SAR aufgezeigt. Eine besondere Herausforderung und gleichzeitig ein zentraler Aspekt dieser Arbeit ist der Einsatz von SAR-Tomographie bei großen Aspekt Winkeln. Da über größere Winkelbereiche die Rückstreueigenschaften dieser Objekte oftmals stark fluktuierend sind, ist eine Anpassung der klassischen Methoden notwendig, um verlässliche Resultate zu erzielen. Zur Validierung der verwendeten Algorithmen werden Ergebnisse des vom DLR entwickelten Radarsystems TIRAMI-SAR verwendet.
Alexander Haas "Charakterisierung von Gebäudestrukturen mit Mikrowellenmessverfahren"
Dienstag, 21. Juli 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Das erklärte Ziel der deutschen Bundesregierung ist es, den Primärenergiebedarf der Bundesrepublik Deutschland bis zum Jahr 2050 auf 50 % des Werts aus dem Jahr 2008 zu senken. Um dieses Ziel zu erreichen, muss der Verbrauch in allen Sektoren – Industrie, Verkehr, Haushalte etc. – reduziert werden. Der Gebäudesektor macht zwecks Erzeugung von Raumwärme einen Anteil von 26 % am Endenergieverbrauch in Deutschland aus. Einsparungen in diesem Sektor könnten insbesondere durch Dämmsysteme erreicht werden. Um deren Potenzial optimal ausnutzen zu können, bedarf es jedoch präziser Informationen über den Aufbau der zu dämmenden Wände. Dies stellt insofern eine Herausforderung dar, als ein Großteil des Gebäudebestands in Deutschland auf Altbauwohnungen entfällt, die vor oder kurz nach dem Ende des Zweiten Weltkriegs gebaut wurden. Für einen Großteil dieser Gebäude existieren keine oder nur unzureichende Informationen über den tatsächlichen Wandaufbau. Um diesem Problem anhand präziser Analysen Abhilfe zu schaffen, wird im vorliegenden Vortrag eine Gesamtmethodik mittels Mikrowellenradar zur Wandcharakterisierung entwickelt. Hierzu wurden erstmals Untersuchungen an mehrschichtigen Wandaufbauten unterschiedlicher Zusammensetzung durchgeführt. Aus Reflexionsmessungen an der Gebäudehülle werden Echobilder der Wand erzeugt, aus denen Rückschlüsse auf die hinsichtlich einer optimalen Dämmung relevanten Parameter Schichtanzahl, Schichtdicke und Schichtmaterialien der Wand gezogen werden können. Die Forschungsergebnisse dieser Arbeit dienen als Grundlage für zukünftige, schnell operierende Radarmesssysteme (z.B. auf Flugdrohnen), mit denen in urbanen Umgebungen in Form von Reflexionsmessungen ein Maximum an Informationen über die innere Struktur des jeweils vorhandenen Gebäudebestands extrahiert werden kann.
Carlos Villamil Lopez "Change detection for monitoring of man-made objects using time series of very high resolution spaceborne SAR images"
Mittwoch, 12. Juli 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Earth observation satellites with very high resolution (VHR) capabilities can be exploited to monitor different types of human activity, delivering insights which can be used to inform policy- and decision-making. Synthetic aperture radar (SAR) sensors are especially interesting because they can provide VHR images independently of sunlight and under all weather conditions, and are also very well suited for change detection. This thesis presents three novel methods for the monitoring of man-made objects using time series of VHR SAR images. These methods are then applied to time series of TerraSAR-X images with submeter resolution to address different practical applications: the monitoring of oil inventories, the monitoring of construction activity, and the monitoring of airport activity. The analysis of the obtained results has shown that all three methods perform well, indicating that they could be used in practice for the intended applications.
Jakob Ludwig "Uncertainty Estimation of Convolutional Networks Segmenting Flooded Areas in SAR Imagery"
Dienstag, 9. Januar 2024 um 14.00 h
Dieser Vortrag findet hybrid statt!
Muss leider verschoben werden. Der neue Termin wird zeitnah bekanntgegeben.
Abstract:
Convolutional networks are typically the state-of-the-art for image analysis, but are prone to be overconfident with respect to their predictions. This talk addresses a variety of uncertainty estimation methods applicable to already trained models. The focus is in particular on student ensembles, a post-hoc variation of the epistemic deep ensembles uncertainty quantification based on knowledge distillation. Detailed experiments in the context of flood prediction from SAR imagery leveraging a universal toolbox compare a variety of established uncertainty approaches and quantification metrics.
Nikita Basargin “SAR Data Tensor Decomposition for Bio- and Geophysical Parameter Estimation"
Dienstag, 04. Juli 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Synthetic Aperture Radar (SAR) sensors provide data in polarimetric, interferometric, temporal, and spatial dimensions depending on the acquisition setup. With the increasing availability of multidimensional SAR data the joint processing and information extraction from several data dimensions grows in relevance. We introduce a decomposition framework for multidimensional SAR data that allows an arbitrary number of data dimensions and is based on the Canonical Polyadic (CP) tensor decomposition. The decomposition is formulated as an optimization problem allowing precise control over the shape and properties of the output factors. The specifics of SAR data are taken into account by adding constraints for physical validity and interpretability. In order to demonstrate the framework, a decomposition for polarimetric time series is formulated. The algorithm decomposes a stack of polarimetric coherency matrices into several components. Each component is defined by a polarimetric and a temporal factor that describe the changes in the time series in a compact way. The results are visualized using the polarimetric change matrices and show additional fine-grained changes in comparison to the original change analysis method. Furthermore, we discuss the potential for bio- and geophysical parameter estimation by integrating physical models into the decomposition framework.
David Thomsu “Exploitation of 2-look ScanSAR with ROSE-L for along-track surface deformation measurements"
Montag, 12. Juni 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
In this master's thesis, the concept of 2-look ScanSAR in the context of the ROSE-L mission is investigated in order to retrieve a highly accurate estimate of the along-track ground surface deformation. The novel addition to the concept is the exploitation of the full available azimuth bandwidth in order to increase the overlap between bursts to be able to exploit the spectral diversity directly from the burst data without modifying the nominal ROSE-L acquisition timeline, i.e., the azimuth resolution is preserved. In order to perform this investigation, an interferometric simulator has been implemented, which includes the generation of multi-channel (in azimuth) single-look complex raw data for distributed scenes, followed by the azimuth reconstruction and SAR processing steps. By using this simulator, the quality of the along-track shift estimation is investigated quantitatively for a couple of representative test scenarios. The ROSE-L mission is an extension of ESA's Copernicus program for Earth observation. It is designed to operate as a complement to the Sentinel-1 mission by performing radar observations at L band. One of the goals is to investigate geohazards related to ground surface motion, such as earthquakes, urban subsidence, landslides and flooding, for which an accurate measurement of the azimuth (along-track) deformation would be very valuable. The investigations conducted during this master's thesis show that such measurements are indeed possible, while at the same time a dedicated interferometric processing chain for ROSE-L is proposed in order to perform such measurements.
Simone De Palma “Observation requirements for ionospheric tomographic reconstructions with polarimetric SAR acquisitions”
Donnerstag, 20. April 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Spaceborne synthetic aperture radar (SAR) acquisitions are distorted by the propagation of the transmitted and backscattered pulses through the ionosphere especially at low frequencies. The knowledge of the free electron concentration in the slant range direction in terms of (slant) total electron content (TEC) is critical to correct the amplitude, phase and polarization distortions in SAR images. TEC values can be obtained from precise measurements of Faraday rotation, enabled by the availability of fully polarimetric SAR data. However, an accurate conversion of Faraday rotation to TEC requires the knowledge of the (slant) electron density profiles. Conventional ionosphere mapping techniques (like ionosondes, incoherent ionospheric scatter radars or GPS occultation measurements) can not provide these profiles with the necessary spatial/temporal sampling and resolution for an operational use. As a consequence, evaluating the possibility to obtain them from the SAR data themselves is an important issue. Tomographic reconstructions of the electron density profiles in the plane defined by the orbit and the slant range direction can be obtained from Faraday rotation measurements in multiple azimuth sub-bands. After gridding the orbit-slant range plane, each grid cell is mapped into the sub-bands with a different line-of sight (LOS). In this way, a linear system of equations linking the unknown electron density in each grid cell and the sub-band Faraday rotation measurements can be obtained. The system coefficients depend on the variation of the geomagnetic field component parallel to the LOS within the synthetic aperture, and determine the sensitivity of the measurements to the electron density profiles and in turn the accuracy of the tomographic reconstructions. In this work, this concept for ionospheric tomographic reconstructions is further explored to understand which observation conditions increase the sensitivity of the sub-band Faraday rotation measurements to the spatial variability of electron density. The role of the synthetic aperture length (determined by the azimuth antenna aperture and the off-nadir angle) in providing the necessary LOS diversity is addressed. Further, the effect of geomagnetic field variations within the aperture is considered. Indeed, moving towards equatorial geometries the geomagnetic field becomes more and more orthogonal to the LOS and its variation across the sub-bands increases. Finally, the effect of the characteristics of the underlying ionosphere is investigated, by contrasting situations with irregularities (which favour the reconstruction) to more homogeneous ones. The analysis is carried out by means of simulated data in which simplified profiles (e.g. a Chapman layer) are assumed, and by processing real ALOS-PALSAR and ALOS-2 fully polarimetric data. Although preliminary, the results are expected to be relevant not only for the upcoming ESA BIOMASS mission, but even to inspire new SAR mission concepts for ionosphere mapping.
Daniel Seyer Lavandeira “Investigation of Ocean Clutter Suppression Methods for Improving the Detectability of Ships in Range-Compressed Multi-Channel Airborne Radar Data”
Dienstag, 18. April 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
The water surfaces of our blue planet host numerous human-related activities, making surveillance and monitoring of them a major security concern for many nations. Despite many advanced surveillance systems, safety and security cannot be fully ensured on open seas, as the detection of maritime threats remains a challenging task. Airborne radars offer a great opportunity to complement these systems by filling information gaps through increased revisiting rates and extended observation times in all-weather conditions, provided during both, day, and night times. For this reason, DLR is currently developing a high-altitude platform (HAP), setting out to integrate a novel radar-based maritime surveillance processing chain. A major component in the processing chain is the target detection, which is strongly limited by the presence of sea clutter. Thus, to improve target detection, as well as to reduce the false alarm rate caused by sea clutter, the suppression of unwanted sea clutter is indispensable. This bachelor thesis presents new clutter suppression methods using multichannel airborne radar data to suppress the sea clutter. The radar data used for the investigations are range-compressed (RC) data, as they do not require computationally-heavy synthetic aperture radar (SAR) processing. This saves substantial processing time, enabling real-time operability. Moreover, the clutter suppression is carried out in the range-Doppler domain of the RC data, allowing the integration into the maritime surveillance processing chain, which entirely operates in the Doppler domain. The clutter suppression methods developed in this thesis are based on the displaced phase center antenna (DPCA) technique. Albeit generally applied on land clutter and fully focused SAR images, this thesis explores the application of the DPCA method on sea clutter. Before implementing the DPCA method, data calibration is carried out as a pre-processing step to analyse the resulting effects on the DPCA-based suppression performance. This is followed by expanding the state-of-the-art DPCA method, limited to two receiving channels, to work with multiple channels in a cascaded manner, exploring the possibility of further sea clutter suppression. Furthermore, as the suppression performance may be degraded by the sea surface motion, the DPCA method is conflated with the ATI (along-track interferometry) technique, allowing for a compensation of the sea surface motion by computing the ATI phase. Finally, the aforementioned suppression methods are probed on real marine targets, evaluating their impact on the target detectability.
Keith Morrison “Soil Moisture and Soil Depth Retrieval Using the Coupled Phase-Amplitude Behavior of C-Band Radar Backscatter in the Presence of Sub-Surface Scattering”
Dienstag, 14. März 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
In low-moisture regimes, strongly-reflecting bedrock underlying a soil could provide a dominant return. This offer a novel opportunity to retrieve both the volumetric moisture fraction (mv) and depth (d) of a soil layer using differential SAR interferometry (DInSAR). A radar wave traversing the overlying soil slows in response to moisture state; moisture dynamics are thus recorded as variations in travel time - captured back at a radar platform as changes in phase. The Phase Scaled Dielectric (PSD) model introduced here converts phase changes to those in soil dielectric as an intermediate step to estimating mv. Simulations utilizing a real soil moisture time series from an arid site in Sudan were used to demonstrate the linked behaviors of the soil and radar variables, and detail the PSD principle. A laboratory validation used a soil with a wet top layer variable in depth 1-2 cm and drying from mv~0.2 m3m-3, overlying a gravel layer at a depth of 11 cm. The scheme retrieved d=1.49 ± 0.33 cm and a change Δmv = 0.191-0.021 ± 0.009 m3m-3. The PSD scheme outlined here opens up a new avenue for the diagnostic measurement of soil parameters which is not currently available to radar remote sensing.
Daniel Carcereri “A deep-learning based approach for large-scale forest height estimation”
Dienstag, 07. März 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Large-scale and up-to-date canopy height model (CHM) estimates are of key importance for forest resources monitoring and disturbance analysis. In this seminar we present a study on the potential of Deep Learning (DL) for the regression of forest height from TanDEM-X bistatic interferometric (InSAR) data. We propose a novel fully convolutional neural network (CNN) framework, trained in a supervised manner using reference CHM measurements derived from the LiDAR LVIS airborne sensor from NASA. The reference measurements were acquired during the joint NASA-ESA 2016 AfriSAR campaign over five tropical sites in Gabon, Africa. Together with the DL architecture and training strategy, we present a series of experiments to assess the impact of different input features on the network estimation accuracy (in particular of bistatic InSAR-related ones). Furthermore, we perform a spatial transfer analysis aimed at deriving first insights on the generalization capability of the network when trained and tested on data sets acquired over different locations, combining different kinds of tropical vegetation. The obtained results are extremely promising and already in line with state-of-the-art methods based on both physical-based modelling and data-driven approaches, with the remarkable advantage of requiring only one single TanDEM-X acquisition at inference time.
Julian Lenhart “Precise alignment determination of corner reflectors with GNSS compass and inclinometer”
Dienstag, 24. Januar 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
A precise alignment of corner reflectors is a crucial factor for a correct calibration of satellite based radar systems. Deviations from a perfect alignment directly result in errors in the calibration. To avoid misalignment, a precise measurement of the alignment of corner reflectors is needed. In this master thesis, the concept, development and validation of a measurement instrument based on a GNSS receiver and a two-axis inclinometer for precise alignment determination for corner reflectors is presented. The device uses a configuration of multiple GNSS antennas for the determination of the azimuth angle and an inclinometer for the determination of the elevation and roll angle with an uncertainty below 0.5°. Additionally, the instrument calculates the phase center position of the corner reflector with GNSS real time kinematic (RTK) method with an uncertainty up to a few millimeters from the known corner geometry and stores the results in an internal database. The alignment and the position can be viewed in real time on a mobile phone allowing comfortable and time efficient monitoring of the manual alignment process. Different evaluation methods and the errors influencing the measurement are analyzed and a suitable verification strategy is selected. The uncertainty estimation and comprehensive validation are presented as well.
Dominik Gerber “Road Surface Roughness Estimation With Spaceborne Synthetic Aperture Radar”
Donnerstag, 12. Januar 2023 um 14.00 h
Dieser Vortrag findet virtuell statt!
Abstract:
Road surface roughness is a crucial factor in road traffic safety. It has a direct connection to the grip and skid resistance of vehicles, which influences their acceleration, maneuverability and braking behavior. Therefore, the quality of road roughness must be checked regularly to ensure that the values are within the acceptable range. Until now, road monitoring in Germany has been carried out by measuring vehicles that drive over the highways approximately once in every 4 years. However, this procedure is costly and requires a lot of time and personnel. The German Aerospace Center (DLR) has proved that high resolution airborne X-band synthetic aperture radar (SAR) datasets can be used to estimate the road surface roughness over a wide scale. But, the area that can be covered by an airborne SAR system is still limited. This limitation can be solved by the use of spaceborne SAR. In the scope of this master thesis, the potential to estimate the road surface roughness using the high-resolution X-band SAR datasets acquired by the TerraSAR X radar satellite of Germany was evaluated. The semi-empirical surface roughness estimation model developed at DLR for the airborne SAR was used as a foundation for this study. However, compared to the airborne SAR, the distances to the ground are tremendous in the spaceborne SAR case and thus the model was adapted. In addition to modifying the surface roughness model, this study also tested whether the results are credible, as the spaceborne SAR data has more noise, while only a low signal is received from the roads. The surface roughness results estimated using the spaceborne SAR datasets showed a good agreement with the airborne SAR based results and also with the ground truth surface roughness data. For the end user, KML files were generated to visualize the surface roughness images in Google Earth. The results may serve for future research in spaceborne SAR based surface roughness estimation with different settings and from different platforms.
Copyright © 2024 Deutsches Zentrum für Luft- und Raumfahrt e.V. (DLR). Alle Rechte vorbehalten.