The significant increase in the amount of satellite data in recent years along with the increase in computing resources has opened up new possibilities for Earth Observation (EO) data analysis. However, the main obstacle to obtaining highquality products is the limited amount of labeled data. Among the methods that can help solve this type of problem, transfer learning (TL) has gained enormous interest in recent years. It involves using a model trained on one problem to solve another problem. This article reviews TL methods as applied to EO data analyses. It provides a structured overview of methods in relation to applications, publication statistics, and summarizes the advantages and disadvantages noted of previous research.
16 March 2024 in Journal of Geophysical Research: Planets
Aeolian coarse‐grained ripples have been found in all regions investigated by Mars rovers: Meridiani Planum, Gusev crater, Gale crater, and Jezero crater. Therefore, it can be assumed that coarse‐grained ripples are one of the most common landforms on Mars. Studying their formation and evolution gives us the opportunity to determine past and current wind patterns. They are also crucial for understanding the formation and evolution of larger aeolian bedforms. Of all locations studied in situ on Mars, coarse‐grained ripples in extensive (∼100 km2) ripple fields were found only on Meridiani Planum. As coarse‐grained ripples on Mars are not well characterized in the literature, in this work, the morphometry, morphology, spatial distribution, and orientation of coarse‐grained ripples investigated along the 45 km long traverse of the Opportunity rover were analyzed. The obtained results allowed for a more precise definition of coarse‐grained ripples and for distinguishing three classes of coarse‐grained ripples on Meridiani Planum: small, medium, and large. The coarse‐grained ripple activity on Meridiani Planum is now limited due to low material supply, and the relatively strong induration of the ripple surfaces. Even though most of the coarse‐grained ripples on Meridiani Planum were formed thousands of years ago, some smaller coarse‐grained ripples were formed by modern winds.
21 November 2022 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
The aim of the research was to analyze the possibilities of using deep learning methods for classifying multi-source image data for Mars. It should be emphasized that the main goal of the research was to develop a methodology for integrating image data acquired from orbiters (MRO mission's HIRISE camera) and in situ (Opportunity rover's NAVCAM camera) and to use their combined analytical potential. We used a VGG-16-based network for this study, which is well-characterized in the literature and has been successfully applied in a wide range of applications. The article proposes a methodology for the supervised classification of landforms on Mars. The proposed solution was evaluated using the Meridiani Planum area, utilizing neural network deep learning and was based on multi-source image data. We found that our approach classified aeolian reliefs correctly for more than 94% of the test dataset. The classification accuracy increased to almost 96% when using panoramas developed from Opportunity's images and the derivatives of the digital terrain models used during the classification process. It is possible to broaden the proposed concept of multi-source classification and the customized deep learning system to the analysis of other regions of Mars and to multispectral imaging without losing the generalizability of the solution.
The development and popularization of navigation applications are increasing expectations for their quality and functionality. Users need continuous navigation not only outdoors, but also indoors. In this case, however, the perception of space and movement is somewhat different than it is outside. One potential method of meeting this need may be the use of so-called geo-descriptions—multi-level textual descriptions relating to a point, line or area in a building. Currently, geo-descriptions are created manually. However, this is a rather time-consuming and complex process. Therefore, this study undertook to automate this process as much as possible. The study uses classical methods of spatial analysis from GIS systems and text generation methods based on artificial intelligence (AI) techniques, i.e., large language models (LLM). In this article, special attention will be paid to the second group of methods. As part of the first stage of the research, which was aimed at testing the proposed concept, the possibility of LLMs creating a natural description of space based on a list of features of a given place obtained by other methods (input parameters for AI), such as coordinates and categories of rooms around a given point, etc., was tested. The focus is on interior spaces and a few selected features of a particular place. In the next stages, it is planned to extend the research to spaces outside buildings. In addition, artificial intelligence can be used to provide the input parameters mentioned above.
Kampinoski Park Narodowy nie prowadzi na swoim terenie stałego monitoringu ruchu turystycznego. Nie jest tym samym znana liczba odwiedzających, a także wzorce ich przestrzennych zachowań, w tym również takich, które mogą naruszać obowiązujące przepisy w zakresie ochrony przyrody. Celem pracy była analiza aktywności użytkowników portalu STRAVA w latach 2019‑2023 oraz określenie wielkości ruchu poza wyznaczonymi do tego celu obiektami liniowymi. Analizie poddano wszystkie obiekty liniowe (udostępnione i nieudostępnione do ruchu turystycznego) znajdujące się w bazie OpenStreetMap na obszarze Kampinoskiego Parku Narodowego. Kwantyfikacja obiektów liniowych pod kątem intensywności wykorzystania wykazała, że bez względu na rodzaj aktywności (spacerowanie, bieganie, jazda na rowerze) najbardziej intensywnie wykorzystywanymi obiektami liniowymi w parku są te położone w jego wschodniej części, graniczącej z m.st. Warszawą. Najwyższy poziom aktywności (10 decyl, oznaczający liczbę aktywności z przedziału 15 061‑88 305) zaobserwowano na 839 odcinkach tras, z tego 4,9% stanowiły odcinki nieudostępnione dla ruchu. Przestrzenne wzorce są inne dla rowerzystów a inne dla spacerowiczów i biegaczy. Opracowana na podstawie rzeczywistych danych o użytkowaniu poszczególnych obiektów mapa intensywności aktywności może stanowić istotny element wspierania decyzji podejmowanych w zagadnieniach związanych z udostępnianiem parku społeczeństwu. Praca stanowi pierwsze tego typu podejście do skwantyfikowania intensywności ruchu turystycznego w parku narodowym w oparciu o obiektywne dane liczbowe.
The research aims to create a tool that, based on a loaded point cloud from airborne laser scanning (LIDAR), allows a reading of the heights of various objects, especially buildings, and an easy calculation of height differences. Usually, tools of this type operate on full 3D models and require advanced applications such as GIS or CAD class. In this case, it will be possible to analyze the point cloud in a 'perspective' (3D) view in conjunction with other data. The service, combined with the possibility of attractive geovisualisation, in the background of various geospatial data, e.g. aerial photographs, will facilitate geospatial analyses necessary for smart cities. It will be possible to work on integrated, large sets of point clouds. Also, the possibility of automatic detection of the type of indicated object in an automated manner, with the use of the Warsaw University of Technology's proprietary software library CENAGIS.SAND and the use of AI algorithms for this purpose, will be transferred. Advances in AI show the possibility of creating classification algorithms that work on image data, like visualization of the point cloud, and on point cloud itself. During the experiment, it is necessary to analyze possible AI solutions to efficiently and accurately classify point cloud datasets and return the leading type of development and leading type of land cover on the analyzed area. The desired level of technology readiness is a fully automated method that can take point cloud and area as input, and return information about the leading type of development and leading type of land cover as a result.