Prototypes + Software

Eigenmaps (Web-based data visualization and querying) – Simplefleet (floating car data-based services) – Jeocrowd (Browser-based computing) – Map Construction – Map Matching – CITER (history textbook browser) – Geoblogging (map-based travel narratives) – Archaiorama (archaological GIS) – Spatiotemporal Indexing


Eigenmaps – Data Visualization and Querying platform

The Eigenmaps platform is a data visualization and querying platform for unstructured (textual) data. As we see from daily experience, exploratory data analysis is an important but also challenging task when trying to make sense of datasets. Many times, visualizing the data in different ways creates a first insight into the data. Eingemaps allows us not only to visualize, but to query the data using the very visualizations. Using spatial, temporal, thematic metadata in combination with a social communications graph, we are able to select spatial ranges, temporal ranges, certain topics, and users (social aspect) in these visualizations to query the underlying content and identify actionable intelligence in the underlying data. We have used the platform for exploring social media data in the form of tweets, emails and textual documents such as Founding Father Papers, the correspondence of six major shapers of the United States: George Washington, Benjamin Franklin, John Adams (and family), Thomas Jefferson, Alexander Hamilton, and James Madison, which include 180,000 documents.  

The platform utilizes a scalable data management architecture (Mongo and Elasticsearch) in connection with a Javascript Web frontend.

Maps.SimpleFleet – Isochronoes and Travel Time Visualization



Dashboard interface

System architecture


GPS positioning devices are becoming a commodity sensor platform with the emergence and popularity of smartphones. This abundance of usually low-sampling-rate (e.g., one point every 1-5 minutes) GPS trajectories have lead to significant increase in research activities around map-matching, the process of aligning a sequence of observed user traces to the underlying road network graph. Still at this moment, practical uses of this research have only been considered in costly and complex fleet management applica- tions. The SimpleFleet platform tries to address these shortcomings by creating an efficient infrastructure for low-cost fleet management solutions. Right now, the resulting SimpleFleet service is fully operational for three European cities namely: Athens (Greece), Berlin (Germany) and Vienna (Austria).

The SimpleFleet demonstrator provides an interactive Web interface based on a “slippy map” with area buttons at the bottom for switching between the three available cities. For each city, the demo services the following map layers:

  • Live-traffic – actual travel times updated every 5 minutes
  • Speed profiles – typical speeds for a given hour of the day computed from several months of tracking data
  • Traffic message channel alerts (TMCs) – for Berlin only. TMC alerts are short informative messages which appear in the electronic road signs above major roads.
  • Isochrones – for a given location, the default setting returns six isochrones using 5 minute travel time intervals up to 30 minutes.
  • Routing – shortest-path computation that considers live traffic data and historic speed profiles. Calculates a path along with the expected travel time and distance based on the current and expected speeds.


Maps.SimpleFleet – Isochronoes and Travel Time Visualization



Shortest path computation

System overview

JEOCROWD Collaborative Search Engine

Geospatial data is out there, maybe not always in the format, quality and amount we expect it, but by employing intelligent data collection and mining algorithms, we are able to produce high quality, and what is sometimes more important, unusual data. In this context, our contribution exploits already existing data and employs user-contributed computing resources and search parameters to produce quality geospatial datasets. We refer to this approach as collaborative geospatial feature search.

The Web has created a number of services that facilitate the collection of location-relevant content, i.e., data for which location is just but one, and most often not the most important attribute. Prominent examples include (i) photo sharing sites such as Flickr and Panoramio, and many others, (ii) microblogging sites with a geotagging feature, e.g., Facebook, Google+, Twitter as well as related photo sharing sites (twitpic) and (iii) more recently check-in services, e.g., Foursquare, Loopt, and Gowalla. Our goal is to utilize this user-contributed geospatial data and to derive meaningful geospatial datasets from it. This work advocates crowdsourcing as a means to mine such user-generated point-cloud data. It employs browser-based collaborative search for deriving the extents of geospatial objects, or, Points of Interest from point-cloud data such as Flickr image locations. In a nutshell, the user provides the search terms (e.g., “Plaka, Athens”) and respective point databases are queried based on their tag information and using their APIs to retrieve a point cloud that characterizes this location. This point cloud stands for the wisdom of the crowd and somehow this data needs to be aggregated to derive the actual location of the searched-for geospatial object, e.g., a polygon derived from the point cloud and representing the spatial extent of “Plaka”.

Here, a particularity in our approach is the use of a browser as a computing platform. Borrowing from the MapReduce computing paradigm, we create our own browser-based version in which data collection and computation tasks are delegated to the connected search clients (browsers) and the server only coordinating the search. The search is collaborative in that the more clients are connected, i.e., users searching for the same term, the faster the search finishes. Overall, the objective of this work is to transform available user-contributed geocontent (“point cloud”) into meaningful chunks of information (e.g., a polygon representing the location of a spatial object). We aim for geospatial datasets obtained with simplicity and speed comparable to that of Web- based search. To achieve this task, we propose our search method utilizing crowdsourcing concepts implemented as a Web-based, collaborative search tool termed Jeocrowd.


Geospatial feature search using Flickr point-cloud data – searching Plaka area in Athens, Greece

Map Construction Algorithms

Road networks are important datasets for an increasing number of applications. However, the creation and maintenance of such datasets pose interesting research challenges. Hence, an automatic road network generation algorithm that takes vehicle tracking data in the form of trajectories as input and produces a road network graph would be beneficial for various scenarios. Our work towards this goal introduces novel road network generation algorithms and provides also the means to assess the quality of the result. The specific approach uses trajectories as a data source. They represent sampled movement at regular intervals using positioning technology such as GPS. The resulting dataset comprises movement trajectories such as the one shown in the figure below. This example dataset comprises vehicle trajectories recorded by means of GPS at a typical sampling rate of 30s (2Hz). The figure on the right shows a road network that was derived from this tracking data. The algorithm producing this result exploits geometric properties of the data to actually merge the geometries of trajectories represented as polylines. Datasets, map construction algorithms, quality assessment algorithms as well as visualization will be made available at as they become available.



Input – vehicle trajectories

Output – a road network graph

Map-Matching Algorithm

Tracking data is obtained by sampling the positions using typically GPS to produce data that in database terms is commonly referred to as trajectories. Unfortunately, this data is not precise due to the measurement error caused by the limited GPS accuracy, and the sampling error caused by the sampling rate, i.e., not knowing where the moving object was in between position samples. A processing step is needed that matches tracking data to the road network. This technique is commonly referred to as map matching.



Measurement and Sampling error

Map-matching example

Unfortunately, this data is not precise due to the measurement error caused by the limited GPS accuracy, and the sampling error caused by the sampling rate. A pre-processing step that matches the trajectories to the road network is needed. This technique is commonly referred to as map matching. Most map-matching algorithms are tailored towards mapping current positions onto a vector representation of a road network. In the given context, the entire trajectory given as a sequence of historic position samples needs to be mapped. The fundamental difference in these two approaches is the error associated with the data. Whereas the data in the former case is mostly affected by the measurement error, the latter case is mostly concerned with the sampling error. We introduced a map-matching algorithm that maps a trajectory onto a road network by matching geometries. This approach aims for a global match mapping of the entire trajectory to a candidate curve in the road network, i.e., find the path in the road network that most closely resembles trajectory. Different similarity measures are used, the Fréchet distance and the weak Fréchet distance, resulting in two different map-matching algorithms, which guarantee to find a matching curve with optimal distance to the trajectory. The essential decision problem is illustrated in the animation below, i.e., find a path in the road network that is within an ε distance of a given trajectory.


Citer – A history textbook browser

The main objective of this effort/project was the creation of a tool to provide unified access to a multi-language digital European history textbook base, which integrates already existing digital material from various European publishers. The integration is achieved by introducing largely language independent metadata including (i) space (locations), (ii) time (dates), and (iii) history concepts. These three metadata aspects represent essential abstractions of the content and are used to index and access the textbook content in a largely language-independent way. The specific content used in this work comprises 57 history textbooks from a total of eight publishers, from seven countries, representing six languages (English, Spanish, German, Italian, Greek and Slovenian).


CITER interface – browsing textbooks using spatial, temporal and a history ontology

Geoblogging Platform

When coming home from a memorable journey, wouldn’t it be great to create a digital replica of the trip, i.e., quickly organize collected images, videos, etc. and have a simple means of adding some thoughts?

With the Geocrowd platform we propose geoblogging as a means for spatiotemporal storytelling, more specifically the story of a journey, be it an afternoon walk in your neighborhood, a chase for a great coffeeshop, your mountainbike trip, or hiking adventure. The scope is to provide a simple to use application that allows one to tell the story of the trip based on the content collected during the trip. The role of content is to support the story. In our application, the essential aspects are a map, a storyboard and a timeline.


Geoblogging platform

ARCHAIORAMA Excavation Documentation System

Excavation documentation is an intriguing subject when confronted with the vastly varying needs of the users, the archaeologists. The basic requirements to a tool supporting such a task are the holistic comprehension, management and promotion of the results of excavations by providing (i) flexible import of data, i.e., location-independent entering of data through Web clients, (ii) dynamic visualization of the excavation space, i.e., use of query functionality to visualize the collected information in 2D and 3D, and (iii) reporting and documentation of the excavational progress by means of printed and electronic reports.

Based on these requirements the prototypical excavation documentation system Archaiorama was developed. It provides a graphical user interface for data collection, data analysis, and reporting and is built on top of a PostgreSQL DBMS implementing a data model specifically designed for this application context. Arxaiorama exhibits a strong spatial data management component and, given its query functionality, visualization capabilities and data model, can be considered a custom geographic information system for excavation documentation.


Archaiorama Excavation Documentation System – functionality

TB-tree and STR-tree Spatiotemporal Access Methods

The following spatiotemporal access methods are available for download:

The code is based on the R*-tree implementation of Bernhard Seeger (SIGMOD 1990).