FIG Peer Review Journal


Multitemporal Data Registration through Global Matching of Networks of Free-form Curves (3464)

Dimitra Vassilaki, Charalabos Ioannidis and Athanassios Stamos (Greece)
Prof. Charalabos Ioannidis
School of Surveying & Rural Engineering
Laboratory of Photogrammetry
National Technical University of Athens
9, Iroon Polytechniou St.
Corresponding author Prof. Charalabos Ioannidis (email: cioannid[at], tel.: + 30 210 7722686)

[ abstract ] [ handouts ] [ handouts ]

Published on the web 2009-02-16
Received 2008-12-01 / Accepted 2009-02-16
This paper is one of selection of papers published for the FIG Working Week 2009 in Eilat, Israel and has undergone the FIG Peer Review Process.

FIG Working Week 2009
ISBN 978-87-90907-73-0 ISSN 2307-4086


In recent years the frequency of geospatial data collection (optical, SAR, lidar, hyperspectral) has been increased and the methods for surveying and mapping of earth’s surface have been improved technically and economically. Consequently, a large volume of multitemporal data is available, which contains information appropriate for a number of applications, such as temporal change detection, coastline or urban development monitoring, deforestation, etc. For the good use of these data several point-based or feature-based registration techniques have been developed. Recently linear features are gaining interest over points in registration procedures, because man made environment is rich of linear features, linear features can be detected more easily and more reliably than points and matching of linear features is often more reliable than point matching. In this paper a method for multitemporal heterogeneous data registration through global matching of networks of free-from curves based on the Iterative Closest Point (ICP) algorithm is presented. The correspondence of each curve in one dataset with a curve in the other dataset must be established before the application of the ICP algorithm. Six different approaches have been tested; finally, a hybrid method is proposed, which includes the calculation of various values as metrics of the distance of the curves in the two datasets. Next, it is necessary to match all the available pairs of curves simultaneously, because the curves share a common transformation; thus it is not possible to match each pair independently, since each matching would produce a different transformation. Alternative techniques have been tested for the minimization of the necessary total computational time, e.g., by automate improvement of the pre-alignment. The performance of the proposed strategy was tested with simulation and real data. The registration of an orthorectified high resolution satellite image with an old medium scale topographic map is given as an application. The procedure followed is described and the results are presented, which are especially promising, as the registration accuracy is of the same order with the accuracies of the data.
Keywords: registration; matching; iterative closest point algorithm; change detection