Mariana Belgiu

Ontology-Based Classification of Building Types Detected from Airborne Laser Scanning Data

Abstract: Accurate information on urban building types plays a crucial role for urban development, planning, and management. In this paper, we apply Object-Based Image Analysis (OBIA) methods to extract buildings from Airborne Laser Scanner (ALS) data and investigate the possibility of classifying detected buildings into ―Residential/Small Buildings‖, ―Apartment Buildings‖, and ―Industrial and Factory Building‖ classes by means of domain ontology and machine learning techniques. The buildings objects are classified using exclusively the information computed from the ALS data. To select the relevant features for predicting the classes of interest, the Random Forest classifier has been applied. The ontology-based classification yielded convincing results for the ―Residential/Small Buildings‖ class (F-Measure 97.7%), whereas the ―Apartment Buildings‖ and ―Industrial and Factory Buildings‖ classes achieved less accurate results (F-Measure 60% and 51%, respectively).

Towards a framework for agent-based image analysis of remote-sensing data

Object-based image analysis (OBIA) as a paradigm for analysing remotely sensed image data has in many cases led to spatially and thematically improved classification results in comparison to pixel-based approaches. Nevertheless, robust and transferable object-based solutions for automated image analysis capable of analysing sets of images or even large image archives without any human interaction are still rare. A major reason for this lack of robustness and transferability is the high complexity of image contents: Especially in very high resolution (VHR) remote-sensing data with varying imaging conditions or sensor characteristics, the variability of the objects’ properties in these varying images is hardly predictable. The work described in this article builds on so-called rule sets. While earlier work has demonstrated that OBIA rule sets bear a high potential of transferability, they need to be adapted manually, or classification results need to be adjusted manually in a post-processing step. In order to automate these adaptation and adjustment procedures, we investigate the coupling, extension and integration of OBIA with the agent-based paradigm, which is exhaustively investigated in software engineering. The aims of such integration are (a) autonomously adapting rule sets and (b) image objects that can adopt and adjust themselves according to different imaging conditions and sensor characteristics. This article focuses on self-adapting image objects and therefore introduces a framework for agent-based image analysis (ABIA).

Coupling formalized knowledge bases with object-based image analysis

Object-based image analysis (OBIA) is a widely used method for knowledge-based interpretation of very high resolution imagery. It relies on expert knowledge to classify the desired classes from the imagery at hand. The definition of classes is subjective, usually project-specific and not shared with the community. Ontologies as a form of knowledge representation technique are acknowledged as solution to establish and document class definitions independently of an OBIA framework. However, ontologies have not yet been strongly integrated in this image analysis framework. This paper presents a method to automatically integrate ontologies in OBIA. The method has been implemented as a tool to be used with the eCognition® software (Trimble, Sunnyvale, CA, USA). A case study was conducted for classifying the land cover classes defined by the Environment Agency of Austria in the Land Information System Austria (LISA) project using WorldView-2 image. The strength of this approach is the direct integration of ontologies into the OBIA process, which reduces the effort necessary to define the classes for image analysis and simultaneously reduces its subjectivity.

Demography of Twitter Users in the City of London: An Exploratory Spatial Data Analysis Approach

Geolocated tweets are not evenly spread across space, but appear in accumulations. By exploring a collection of 3 months of geolocated tweets for London, this work analyses tweet hotspots and demographic characteristics of the wards where these hotspots appear. The Twitter messages are separated into day-time and night-time tweets to support the assumption about work places and home places of Twitter users. Tweets from users with less than three posts in the investigated time period are eliminated to increase the probability of analysing locals rather than tourists. The first step in the analysis is the identification of tweet hotspots. These hotspots are wards, where increased Twitter activities are taking place, as the population figures would suggest. The subsequent step in the analysis deals with the detection of patterns in the relationship between demographic characteristics of London’s wards and the numbers of tweets. This part of the analysis employs exploratory spatial data analysis for generating hypotheses for an ordinary least squares regression analysis. The contribution of this work is the exploration of representations and analyses for investigating who Twitter users in London are.

Automated near real-time earth observation level 2 product generation for semantic querying

Existing Earth observation (EO) content-based image retrieval (CBIR) systems support human-machine interaction through queries by metadata text information, image-wide summary statistics or either image, object or multi-object examples. No semantic CBIR (SCBIR) system in operating mode has ever been developed by the remote sensing (RS) community. At the same time, no EO dataderived Level 2 prototype product has ever been generated systematically at the ground segment, in contrast with the visionary goal of the Global Earth Observation System of Systems (GEOSS) implementation plan for years 2005-2015. Typical EO Level 2 products include: (i) a multi-spectral (MS) image corrected for atmospheric, adjacency and topographic effects, and (ii) a scene classification map (SCM), encompassing cloud and cloud-shadow quality layers. This work presents an original hybrid (combined deductive and inductive) feedback EO image understanding for semantic querying (EO-IU4SQ) system as a proof-of-concept where systematic multi-source EO big data transformation into Level 2 products is accomplished as a pre-condition for SCBIR, in agreement with the object-based image analysis (OBIA) paradigm and the Quality Assurance Framework for Earth Observation (QA4EO) guidelines. In the hybrid EO-IU4SQ system, statistical model-based/bottom-up/inductive/machine learning-from-data algorithms and physical model-based/top-down/deductive/human-to-machine knowledge transfer approaches are combined with feedback loops to take advantage of the complementary features of each and overcome their shortcomings.

Architecture and prototypical implementation of a semantic querying system for big Earth observation image bases

ABSTRACT Spatiotemporal analytics of multi-source Earth observation (EO) big data is a pre-condition for semantic content-based image retrieval (SCBIR). As a proof of concept, an innovative EO semantic querying (EO-SQ) subsystem was designed and prototypically implemented in series with an EO image understanding (EO-IU) subsystem. The EO-IU subsystem is automatically generating ESA Level 2 products (scene classification map, up to basic land cover units) from optical satellite data. The EO-SQ subsystem comprises a graphical user interface (GUI) and an array database embedded in a client server model. In the array database, all EO images are stored as a space-time data cube together with their Level 2 products generated by the EO-IU subsystem. The GUI allows users to (a) develop a conceptual world model based on a graphically supported query pipeline as a combination of spatial and temporal operators and/or standard algorithms and (b) create, save and share within the client-server architecture complex semantic queries/decision rules, suitable for SCBIR and/or spatiotemporal EO image analytics, consistent with the conceptual world model.

Agent based image analysis (ABIA) - preliminary research results from an implemented framework

Object Based Image Analysis (OBIA) has meanwhile been established as a paradigm for analyzing remotely sensed image data. Although the degree of automation for OBIA methods has increased for several applications, especially in the domain of remote sensing, robust and transferable object-based solutions for automated image analysis of sets of images or even large image archives are still rare. One of the reasons for this lack of robustness and transferability is the high complexity of remote sensing image contents: Especially in Very High Resolution (VHR) remote sensing data, under varying imaging conditions or sensor characteristics, the objects’ properties can vary unpredictably. Although earlier work has demonstrated that OBIA rule sets bear a high potential of transferability these rule sets need to be adapted manually in order to receive acceptable results, or the classification results need to be adjusted manually in a post-processing step. In order to automate these adaptation and adjustment procedures we investigate the coupling, extension and integration of OBIA with the agent-based paradigm, which is exhaustively investigated in software engineering and robotics. The aims of such integration are a) rule sets which can be adapted autonomously according to varying imaging data, and b) image objects which can adapt and adjust themselves in order to best possibly represent the objects of interest in an image. This paper briefly introduces a framework for Agent Based Image Analysis (ABIA) and presents our first research results.

Introduction to “Geographic information science: a multi-disciplinary and multi-paradigmatic discipline”

Geographic Information Science (GISc) is a relatively young discipline whose research body revolves around Geographic Information System (GIS) research, technology, and service (Goodchild 1992). This research field is motivated by the desire to improve the efficiency and usability of geospatial technology and applications. Despite significant scientific accomplishments since its inception, there are still numerous challenges needed to be addressed by the GISc community. These challenges include continued research on topics such as the understanding and efficient utilization of the neogeography phenomenon, the meaningful representation and modeling of complex spatiotemporal phenomena, the 3D building information modeling, and the representation of other nonspatial attributes that may exist in space-time (Goodchild 2010). Comprehensive research agendas for GISc have, for example, been put forward by the University Consortium for Geographic Information Science (UCGIS) and the National Center for Geographic Information and Analysis (NCGIA). The articles included in this special issue “Geographic Information Science: A Multi-Disciplinary and MultiParadigmatic Discipline” investigate a subset of these current research challenges. This special issue illustrates the multidisciplinary research that is currently conducted in the “Doctoral College GIScience,” Interfaculty Department of Geoinformatics – Z_GIS at the Paris-Lodron University of Salzburg, Austria. Financed through the Austrian Science Fund (FWF), this Doctoral College started in fall 2011 for an initial 4-year funding period with the possibility of an extension for another two 4-year terms. Similar to the Integrative Graduate Education and Research Traineeship Program (IGERT) in the US, the Doctoral College funds 10 doctoral students directly through the FWF. Another 10+ associated students are part of the Doctoral College but funded through other sources. The doctoral students are being supervised by a team of 10 core and 16 associated faculty. This faculty team spans across disciplinary boundaries, especially between the geo-disciplines and other natural sciences but also with formalized ties into computing and the social sciences. The research focus of the Doctoral College falls into three interconnected, interdisciplinary research clusters that include “Representations and Data Models”, “Time and Process Models”, and “Spatialization, Media, and Society.” This special issue covers two main sections with three and four articles each and one introductory paper reflecting on the multidisciplinary and multiparadigmatic field of GISc. The first of the two main sections is titled “Representations and Spatial Data Models” and includes three methodological papers that suggest novel ways of spatial data representations and models. The second main section “GIScience Research with User Generated Content and Big Data” summarizes the main contributions of four papers that utilize user generated content and big data and apply them to spatial planning, mobility patterns, and movement measures. The majority of the authors, in most cases the lead author of each of the eight articles, is directly affiliated or has been affiliated with the Department of Geoinformatics – Z_GIS, University of Salzburg. Some of the authors which are not affiliated with the University of Salzburg currently serve as external supervisors of doctoral students in the “Doctoral College GIScience.” Similar to a regular issue of Cartography and Geographic Information Science (CaGIS), all eight articles included in this special issue have undergone a rigorous, double-blind peer-review process. On average, each article was evaluated by 2+ international expert reviewers. In the remainder of this Editorial a brief synopsis of all eight papers will be provided. In their introductory paper, Thomas Blaschke and Helena Merschdorf analyze the research domain of GISc in an attempt to understand the multifaceted scientific nature of this academic field. The authors assess which disciplines contribute to the GISc research body on the basis of peer-reviewed publications available on the ISIWeb of Knowledge and SCOPUS databases. This article revisits the question of whether GISc is a science and emphasizes its multidisciplinary and multiparadigmatic nature. The history, research agenda, and scientific principles of GISc are also thoroughly reviewed.