A5045272531- Remote-Sensing Image Classification
- Remote Sensing and Land Use
- Fire effects on ecosystems
- Fire Detection and Safety Systems
- Spectroscopy and Chemometric Analyses
- Landslides and related hazards
- Superconducting Materials and Applications
- Remote Sensing in Agriculture
- Marine and coastal ecosystems
- Spacecraft and Cryogenic Technologies
- Smart Agriculture and AI
- Water-Energy-Food Nexus Studies
- Marine Biology and Ecology Research
- Advanced Chemical Sensor Technologies
- Oil Spill Detection and Mitigation
- Land Use and Ecosystem Services
- Remote Sensing and LiDAR Applications
- Methane Hydrates and Related Phenomena
- Soil Geostatistics and Mapping
- Cultural Heritage Management and Preservation
- Advanced Image Fusion Techniques
- Arctic and Antarctic ice dynamics
- Oceanographic and Atmospheric Processes
- Coastal and Marine Management
- Identification and Quantification in Food
National Technical University of Athens
2010-2025
Agricultural University of Athens
2023-2024
Institute of Astronomy
2021
National Observatory of Athens
2021
Planetek Hellas (Greece)
2018
National and Kapodistrian University of Athens
2011
The accelerated transition of the maritime industry to low and zero-carbon energy solutions is giving rise a diverse range potential marine fuels, such as liquefied natural gas, ammonia, hydrogen, along with transport temporary storage CO2 from carbon capture units. These substances predominantly require cryogenic handling, which introduces complex system tanks, piping, process equipment onboard ships. core component processes tank, where vapour-liquid equilibrium occurs due external heat...
This paper investigates the applicability of deep learning models for predicting severity forest wildfires, utilizing an innovative benchmark dataset called EO4WildFires. EO4WildFires integrates multispectral imagery from Sentinel-2, SAR data Sentinel-1 and meteorological NASA Power annotated with EFFIS fire detection size estimation. Resulting in a coverage 45 countries (Figure \ref{Figure1-events}) total 31730 wildfire events 2018 to 2022. All these various sources are archived into cubes,...
This paper presents a new nonlinear unmixing method. Based on relative distances which imply nonlinearity, the method introduces “fractional distance” as key variable that quantifies interactions between pixels and endmembers. Relationships fractional abundance fractions are built through networks. Because an equal spectral mixture of ground classes present surface sensed is likely impossible, proposed method, due to its mathematical concept, reveals unknown Three versions have been...
This work introduces a one of its kind labeled Earth Observation based benchmark dataset, Sen4AgriNet, for agricultural applications in Europe. The dataset is using farmer declarations the period 2016–2020, available as open data only very recently. Sen4AgriNet contains 42.5 million parcels hence significantly larger than existing archives and tailored to be used training source context deep learning. It consists two sub-datasets: Object Aggregated Dataset (OAD) Patches Assembled (PAD). OAD...
The paper presents a new method for estimating oil spill thickness using hyperspectral data. relies on the abundance fractions provided by spectral unmixing methods. Given that materials which compose signature of pixels presenting spills are and water, correlation functions between these endmembers were established artificial images. images created laboratory measurements spills. Thirteen different types used production relevant images, each one spread water surface with six levels...
Fronts, which are sharp boundaries between distinct water masses, play a substantial role in managing biodiversity of marine species and preserving resilient ecosystem. The overarching aim this study is to compare different methodologies for detecting fronts. Many fronts identifiable by their strong temperature gradient. For that reason, tests how two edge detection (Laplacian Canny) performs on once applied the Sea Surface Temperature (SST) product Sentinel-3 SLSTR instrument. In few words,...
The aim of the SDGs is to help human activities be sustainable. SDG14 "Oceans" targets at stability and sustainability marine ecosystems their resources. Among its ten targets, 1st refers prevention significant reduction pollution all kinds. To quantify target, 14.1.1 "Index Coastal Eutrophication (ICEP) Floating Plastic Debris Density" introduced by UNEP. Currently, classified in Tier III, i.e. methods data sources for estimation are not defined, whereas type information needed defined. It...
This paper presents a new methodology for the discrimination of plant species and their varieties using hyperspectral data. The concept lies on combination spectral pre-processing algorithms (SPPA) that enhance between varieties. SPPA use as input single signature transform it according to function. A k-step uses k serially. Initially each is used first SPPA. result this second SPPA, so until desired pre-processed signatures are reached. These then discriminated by applying matching...
This paper presents a benchmark dataset called EO4WildFires; multi-sensor (multi spectral; Sentinel-2, Synthetic-Aperture Radar - SAR; Sentinel-1, meteorological parameters; NASA Power) time-series that spans 45 countries, which can be used for developing machine learning and deep methods targeted the estimation of area forest wildfire might cover. novel EO4WildFires is annotated using EFFIS (European Forest Fire Information System) as fire detection size data source. A total 31,730 events...
The landscape, as a comprehensive entity encompassing diverse ecosystems, cultural values, aesthetic qualities, and geological features, is integral to the concept of natural heritage. Climate change poses substantial threat landscapes. impacts are multifaceted, affecting not only ecological integrity these areas but also their significance livelihoods people who depend on them. Efforts mitigate climate adapt its effects crucial preserving valuable landscapes for future generations.TOPIO...
This paper investigates the applicability of deep learning models for predicting severity forest wildfires, utilizing an innovative benchmark dataset called EO4WildFires. EO4WildFires integrates multispectral imagery from Sentinel-2, SAR data Sentinel-1, and meteorological NASA Power annotated with EFFIS fire detection size estimation. These cover 45 countries a total 31,730 wildfire events 2018 to 2022. All these various sources are archived into cubes, intention assessing by considering...
This paper presents a new methodology for automatically labeling extracted endmembers from hyperspectral images, using reference Spectral Library (SL). The SL is structured with three main layers: consisting material, semantic meaning, and illumination conditions in order to classify the endmembers. process based on two similarity metrics: Angle Mapper Cross Correlation. These metrics are initially introduced spectral inequalities which characterize each endmember of SL. During procedure,...
Recently, a new logarithmic mixed pixel classification method has been developed through the establishment of appropriate networks. Based on fact that natural targets do not consist equally distributed components, Network Method (NBM) alerts user for non-sampled endmembers in image scene. In this paper, detection misallocated hyperspectral space is investigated Method. Detection relies misallocation an endmember affects its signature because includes spectral components from other...
This paper presents a new method for automatically determining the optimum regression model, which enable estimation of parameter. The concept lies on combination k spectral pre-processing algorithms (SPPAs) that enhance features correlated to desired Initially algorithm uses as input single signature and transforms it according SPPA function. A k-step SPPAs preprocessing serially. result each is used next SPPA, so until pre-processed signatures are reached. These then three different...
The concept of a relative radiometric normalization (RRN) for multitemporal multispectral and hyperspectral images is investigated. A new automatic RRN methodology presented: the Normalized Proximity Similarity Methodology (NPSM). NPS introduces two spectral metrics, absolute normalized difference (AND) ratio (ANRD), to select Invariant Pixels (IPs). metric also introduced evaluation proposed methodology: root mean square error (RMSE) between values initial subject image relevant...