Presentation
The Geographical Moving objects Data Acquisition (GeoMoviDA) is a set of tools to create movement
data representations from a sequence of satellite or aerial images, for spatiotemporal databases
and GIS.
The aim is to create a continuous (temporally complete) representation about the changes in shape,
orientation and shape of moving objects.
This means that it should be possible to retrieve the current extent and shape of a moving object
for any instant of time, even if the original information about the object moving in the real
world is a series of observations (snapshots) captured at discrete points in time.
The challenge is to create movement representations modeling as closely as possible the
transformation of the objects in the real-world, creating natural morphing sequences from the
source to target shapes and ensuring valid topologies for all intermediate shapes.
In addition, the representation of the moving objects’ transformations must be compatible with the
data models proposed in the spatiotemporal databases literature.
This framework includes a module to extract the objects of interest from a sequence of images, a module to
define the mapping between two polygons representing the shape of a moving object at two consecutive
snapshots, a tool to display the transformation of the moving objects along time, the datasets used
in the experiments and movies putting in evidence some of the outcomes of this work.
The source code and the datasets are publicly available and collaborations for future developments are welcome. Feel free to contact us for additional information.
The contents of this webpage refers to the work submitted to 7th International Congress on
Environmental Modelling and Software (iEMSs, 2014) and so will be maintained for future reference.
Data sources
The case study used in this work concerns the movement of icebergs at the Antartic. The data sets are:
- Ross 2000 (accessed February, 1, 2013): this is a sequence of satellite images tracking Iceberg B-15 during six months in 2000.
These images are avalable at
http://www.atsr.rl.ac.uk/images/sample/ross/index.shtml.
-
Ross 2004
(accessed March, 21, 2014): this is a sequence of images tracking two large blocks (B-15a and B-15j) of the Ross iceberg during
two months in 2004.
-
Atka 2009
(accessed March, 21, 2014): this is a sequence of images tracking a fragment of an iceberg moving at the east part of Atka Bay during
10 days in 2009.
These image sequences are interesting because they involve the modeling of several spatial
transformations, namely translation, rotation, deformation and splitting of objects.
The deformation of the object occurs in all cases.
The predominant movement of the block B15a it is translation while for the block b15-j it is
rotation.
The predominant movement in Ross 2000 is translation and in some cases there is a significant
rotation of the iceberg.
In Atka 2009, translation and rotation are significant in all cases.
Datasets
The datasets extracted from the sequences of images referred in Data sources can be download
here.
Each dataset has two folders and four files.
The folder named Polygons holds a file for each
original image with the coordinates of the vertices that define the shape of the iceberg at each snapshot.
These polygons have been extracted using a semi-automatic procedure, and the same holds for the
correspondences between polygons; hence it is important to make them available to the community to
safeguard the reproducibility of the results in future works.
The folder named Fpoints holds a file with the feature points for each polygon.
The files store the correspondences between polygons in a sequence of observations.
One file gives the correspondences between each pair of consecutive polygons in a sequence of
observations. The other two files give the correspondences considering only the polygons in the
odd and in the even positions of each sequence of observations.
These enable reproducing the sub-sampling approach proposed in this work to evaluate the quality
of the movement data representations.
The last file, named observations, gives the time of the observations in days.
Source code
Currently, GeoMoviDA is a set of standalone tools developed successively by André Oliveira, Luís
Carlos Paulo and André Oliveira.
The source code can be download
here.
Details about these implementation can be found at:
-
Paulo, L, 2012. Morphing techniques in spatiotemporal
databases, Thesis (M.Sc.), Universidade de Aveiro.
-
Mesquita, P., 2013. Morphing Techniques For
Representation of Geographical Moving Objects.
Thesis (M.Sc.), Universidade de Aveiro.
Movies
-
Simple interpolation: this document has two movies displaying the transformation of an iceberg using the approaches B – Perceptually-based and C – Rotation plane.
- Other examples: this document has three movies demonstrating the transformations obtained using other methods that are not covered in the iEMSs 2014 paper due to space restrictions.
GeoMoviDA