ISPRS Scientific Initiative
Summaries of projects funded for 2014
Benchmark on High Density Image Matching for DSM Computation
PI: Norbert Haala, Institute for Photogrammetry, University of Stuttgart
Stereo-matching for the automatic generation of elevation data from aerial images
was originally introduced more than two decades ago. These approaches used
feature based matching algorithms to avoid problems at regions with very limited
texture. Meanwhile, state-of-the-art digital airborne cameras provide imagery of good
dynamic and signal-to-noise-ratio, which supports pixel-wise matching especially for
low-texture areas. Furthermore, recent matching approaches cope with the general
ambiguity of such a per-pixel measurement by introducing additional constraints. Socalled global algorithms explicitly formulate the matching problem under the
assumption of smooth surfaces. This global optimization problem can be
implemented very efficiently using recursive algorithms like scanline optimization. A
very popular and well performing stereo method frequently applied for the processing
of aerial images is semi-global matching (SGM). Also triggered by these
developments, software tools for image based generation of 3D point clouds and
dense Digital Surface Models (DSM) are currently developed by a number of
research institutes and photogrammetric software vendors.
In order to document the progress also in photogrammetric data processing
benchmarks have proven to be extremely useful. Well known examples which
measure the performance of state-of-the-art matching algorithms are the Middlebury
Stereo Vision Page (Scharstein, & Szeliski 2002) or the benchmark on multi-view
stereo reconstruction (Seitz et.al. 2006). These benchmarks provide general purpose
datasets through a platform, where results can be uploaded to automatically yield
quality metrics of the respective approach. While these projects emerged from the
Computer Vision community, the test on the performance of photogrammetric digital
airborne camera systems (Cramer, 2010) was organized by the German society of
Photogrammetry, Remote Sensing and Geoinformation (DGPF).
Also in view of the ongoing software developments, the proposed initiative aims to
evaluate the potential of photogrammetric 3D data capture from automatic image
matching. Basic scope is the evaluation of 3D point clouds and DSM produced from
aerial images with different software systems. Such a comparative evaluation should
provide a platform for software developers to demonstrate the state-of-the-art of their
ongoing developments. Furthermore, it should help potential users like National
Mapping and Cadastral agencies (NMCAs), which consider a state-wide-generation
of high quality DSMs to understand the applicability of such tools while triggering
further developments based on their needs.
TIC – The Tracking and Imaging Challenge 2015
PI: Stephan Winter, University of Melbourne
TIC – the Tracking and Imaging Challenge 2015 is aiming at stimulating research and
excellent work at the interface of two distinct disciplines working with spatial data –
the Computer Vision & Image Processing and the GIScience & spatial computing
community. This is going to be achieved by requiring that two types of data, namely
trajectory data and images, are linked in innovative ways. The challenge is on
purpose broad and open to allow for unorthodox and novel applications and
methods. It is expected that TIC will enhance the visibility of ISPRS research and
events.
ISPRS Benchmark on Urban Object Detection and 3D Building Reconstruction
PI: Markus Gerke, ITC, University of Twente, Enschede
One of the major topics in photogrammetry is the automated extraction of urban
objects from data acquired by airborne sensors. What makes this task challenging is
the very heterogeneous appearance of objects like buildings, streets, trees and cars,
in particular in data having a very high-resolution, which leads to high intra-class
variance while the inter-class variance is low.
It turns out that in this context progress is hampered by the missing availability of
state-of-the-art test data for research, either captured with aerial cameras or airborne
laserscanners. Only few research groups have access to data of highest quality and
resolution and usually these data sets are difficult to be shared between groups due
to copyright regulations etc. Thus, as many groups do not have access to state-ofthe-art data; the majority of scientific publications evaluate novel methods only on a
very limited amount of data, and different groups use different data sets. Results are
therefore hardly comparable, which, we believe, significantly slows down scientific
progress.
The ISPRS WG III/4 benchmark on “Urban Object Detection and 3D Building
Reconstruction” aims at resolving this problem. It offers state-of-the-art data sets
which can be used by interested researchers in order to test their own methods and
algorithms on urban object classification and building reconstruction. Results can be
submitted at any time and are compared to manually annotated 2D and 3D reference
data using a standard evaluation framework.
Despite significant efforts by the working group officials, the preparation of
benchmark data is very expensive in terms of time and money. Although various data
sets have been promised or made available to the WG officials, limited financial
resources severely retard the release of new benchmark data. Funding through the
ISPRS scientific initiative will allow us to heavily increase 2D and 3D reference data
annotation efforts. We believe that this will enable us to at least double the amount of
completely annotated benchmark data until summer 2015.
Status of Global Geospatial Information
PI: Gottfried Konecny, Leibniz University Hannover
The ISPRS WG IV-2 activity is based on an agreement between ISPRS and
UNGGIM signed in December 2011 to cooperate on a survey of the global status of
geospatial information, signed between the 2011 ISPRS President Orhan Altan and
the 2011 Director of UNGGIM Paul Cheung. The ISPRS WG IV-2 was established on
the initiative of the 2012 ISPRS President Chen Jun with the applicants of Leibniz
University Hannover and the UNGGIM Secretariat in New York to be responsible for
designing, issuing and evaluating a questionnaire to the UN member countries on the
status of geospatial information and its infrastructure.
Reports on the preliminary results were given at the UNGGIM 2 Conference in New
York in August 2012, at the ISPRS Congress in Melbourne in August 2012, at the
UNGGIM Forum in Qatar in February 2013, at the UNGGIM3 Conference in
Cambridge in July 2013, at the UNRCC-Americas in New York in August 2013 to
solicit a wide interest of the UN member states in the effort. The WG IV-2 also held
meetings at Interexpo Geosibir in Novosibirsk in April 2013, at the Geospatial Forum
in Rotterdam in May 2013 and at the Scanex Conference in Moscow in October 2013
to address the issues of non-governmental data providers, such as Google, Microsoft
and others. With the exception of Cambridge in July 2013, ISPRS has so far not been
involved in financial support of these efforts.
Based on this activity the 2014 Project applied for, needs to concentrate on a further
in depth analysis of the collected data and to augment the information from other
sources. Supplementary information has been received by UN Country reports and
from other sources.
A new benchmark dataset for multi-platform very high resolution
photogrammetry
PI: Francesco Nex, Fondazione Bruno Kessler (FBK), Trento
Airborne high resolution oblique imagery systems and RPAS/UAVs are very
promising technologies that will keep on influencing the development of geomatics in
the future years closing the gap between terrestrial and classical aerial acquisitions.
These two platforms are also a promising solution for National Mapping and
Cartographic Agencies (NMCA) as they allow to derive complementary and
sometimes better mapping information. Although the interest for the registration and
integration of aerial and terrestrial data is constantly increasing, only few works have
been truly performed on this topic. Several investigations still need to be undertaken
concerning algorithms ability for automatic co-registration, accurate point cloud
generation and feature extraction from multi-platform image data. One of the biggest
obstacles is the non-availability of reliable and free datasets to test and compare
new algorithms and procedures. In particular, a complete benchmark on oblique,
UAV and terrestrial images together has never been realized before.
For these reasons, this scientific initiative will use the grant provided by ISPRS
mainly to collect and share state-of-the-art multi-sensor data over an urban area,
specifically: oblique airborne images, UAV-based images and terrestrial images of
some selected buildings. In addition, reference points and Terrestrial Laser Scanning
(TLS) data will be acquired to enable a thorough evaluation of image orientation and
calibration methods, as well as of dense image matching results. Interested
researchers will be encouraged to use the image data (from one platform, or a
merge) to participate in the orientation benchmark through submitting 3D coordinates
of independent check points distributed all over the area. Results from image dense
matching will be evaluated using cross-sections, best fitting and high-resolution TLS
point clouds. Although the benchmark will focus on the mentioned tasks, other
research, e.g. on radiometric properties or semantic scene analysis using those
images can be accommodated. Through the provision of this dataset to the
photogrammetric, computer vision and adjacent communities, we expect that the
value of those state-of-the-art platform’s data can be evaluated more objectively. Last
not least, the visibility of ISPRS across domains will be promoted and its connection
with other societies (e.g. EuroSDR) and communities will be strengthened.
Development of a Curriculum for The African Geospatial Sciences Institute
(AGSI)
PI: Ian Dowman, University College London
The primary objective of this project is to support capacity building in North Africa.
North Africa has been subject to severe disruption in recent years and there is an
urgent need for the construction of infrastructure and for the services of trained
geospatial professionals and technicians. The African Geospatial Sciences Institute
(AGSI) is a non‐profit organisation registered in Germany which is supported by
industry, government and academia. The objective of AGSI is to accelerate the
geospatial capacity development in North Africa by providing the facilities for
geospatial projects and management training to regional government employees,
university graduates, private individuals and companies.
Funds from the ISPRS Scientific Initiative would be used to develop a curriculum for
AGSI based on the requirements of the region and using the expertise of ISPRS
working groups. ISPRS could make a significant contribution to developing the
curriculum which could then be used in similar initiatives elsewhere. The project
involves ISPRS Working Groups, ISPRS Regional Members, cognate societies and
local organisations.
The project will last for six months and will involve a survey of potential stakeholders
in North Africa, drafting a curriculum on the basis of the returns from the survey and
then discussing this at a workshop to be held in Tunis in May 2014.
The initial curriculum will concentrate on vocational geospatial training, emphasizing
practical, project-based work. Learning-by‐doing is the priority and whilst all
participants will be encouraged to continue with a more formal education, there will
also be support for anyone who wants to start his own business. The curriculum will
take account of the resources available and particularly look at how low cost
equipment such as UAVs and low cost or open source software can be used. The
emphasis will be on training and education which is fit for purpose in the region and
which will allow those completing the course to work efficiently and effectively.
ISPRS will benefit from enhancement of the international standing and from the
opportunity to contribute significantly to the development of geospatial science in
North Africa. The project will also bring together ISPRS members in the region to
work on a substantive regional project.