OOA-group
What is object-oriented analysis (OOA)?
Image analysis has traditionally been primarily based on pixels and their spectral information. Even more recently developed methods, such as those based on texture, advanced classification or multi-data integration, tend to have difficulties to support contextual analysis. The idea of object-oriented analysis (OOA) is that images are broken down into spectrally homogenous segments or objects, whereby the size and shape can be constrained by a number of parameters, and that additional thematic layers, such as of cadastral boundaries or geological units, can be used to create meaningful segments. For all created segments a large number of parameters can be automatically calculated, providing information on all aspects, such as the segment’s spectral characteristics, texture, shape, orientation, proximity or adjacency to other objects, etc. All of those characteristics can then be used to develop rules that are in turn used to classify the segments. As often different spatial scales can exist in an image (for example we might be interested in a single tree or an entire forest stand, or a chimney vs. the entire roof), a multi-scale segmentation is also possible, leading to different segmentation levels. The relationship of a given object to its so-called super- or sub-objects can then also be used. The strength of OOA is that we can thus effectively integrate process and feature knowledge, which is not possible with most other image processing methods. For example, we can automatically distinguish a swimming pool from a natural pond or other water body based on what we know to define either, or even characterize a certain urban neighborhood, such as formal or informal, based on our definition and concept of those urban settlement types. In effect, OOA is well suited to study not only landcover, but also landuse, and lends itself well to distinguish subtle process differences such as between different landslide types by integrating spectral, shape and geomorphometric parameters.
What does ITC’s OOA Group do?
At ITC we have been conducting research in object-oriented analysis since 2006, focusing on a large range of research problems. This reflects that OOA is a very versatile analysis approach that is of use wherever we want to conduct knowledge-driven processing, i.e. where we can specifically characterize features or processes of interest. To date, the group has focused on the following research fields, using OOA:
- Assessment of social vulnerability
- Extraction of urban elements at risk (e.g. buildings) for improved flood modelling
- Information extraction from (hardcopy) legacy population maps
- (Semi-)automatic detection and type characterisation of landslides
- Detection of erosion features (gullies and rill erosion), and characterisation in the context of landuse systems
- Urban structural assessment, including automatic delineation of homogenous urban patches (structurally and formally homogenous areas with clear boundaries), and road network extraction
- Coal fire-related landcover mapping
In addition we currently focus on more objective segmentation, as well as on reducing the data- or threshold-driven processing that has been characterising OOA work, but which limits transferability of the developed rule sets.
The group members and their projects
 |
Dr. Norman Kerle – group leader |
 |
Tapas Martha (PhD candidate) |
 |
Shruthi Rajesh (PhD candidate) Detection and characterisation of erosion features |
 |
André Stumpf (PhD candidate) Multi-scale landslide detection for monitoring and early warning |
 |
Tolga Görüm (PhD candidate) Improved methods for earthquake-induced landslide susceptibility and hazard assessment |
 |
Melanie Harris (MSc candidate) Integration of landslide anatomy and use of texture in landslide detection |
 |
Janak Joshi (MSc candidate) Improvement of photogrammetric digital surface models of dense urban areas with OOA |
 |
Nawaraj Shrestha (MSc candidate) Correction of photogrammetric floodplain digital surface models for improved flood modelling |
OOA-related journal articles
Martha, T.R., Kerle, N., Jetten, V.G., van Westen, C.J. and Vinod Kumar, K. (2009) Characterising spectral, spatial and morphometric properties of landslides for semi - automatic detection using object - oriented methods. In: Geomorphology, in press.
Kerle, N. and de Leeuw, J. (2009) Reviving legacy population maps with object - oriented image processing techniques. In: IEEE Transactions on Geoscience and Remote Sensing, 47 (2009)7, pp. 2392-2404.
Ebert, A., Kerle, N. and Stein, A. (2009) Urban social vulnerability assessment with physical proxies and spatial metrics derived from air- and spaceborne imagery and GIS data. In: Natural Hazards, 48 (2009)2, pp. 275-294.
Yan Gao, Kerle, N. and Mas, J.F. (2009) Object based image analysis for coal fire related land cover mapping coal mining areas. In: Geocarto International, 24 (2009)1, pp. 25-36.
Conference proceedings
Sliuzas, R.V., Kerle, N. and Kuffer, M. (2008) Object - oriented mapping of urban poverty and deprivation. In: Proceedings of the 4th EARSeL workshop on remote sensing for developing countries in conjunction with GISDECO 8, June 4-7, 2008, Istanbul, Turkey / European Association of Remote Sensing Laboratories (EARSeL), 12 p.
Shamaoma, H., Kerle, N. and Alkema, D. (2006) Extraction of flood - modelling related base - data from multi - source remote sensing imagery. In: ISPRS mid-term symposium 2006 remote sensing: from pixels to processes, 8-11 May 2006, Enschede, the Netherlands, 7 p.
eCognition corner
Most of the OOA work of the group is being carried out with Definiens’s eCognition software. In principle developed methods should be transferable to other data types and areas, though this needs for testing, and adaptation will be needed. Here we make available test data sets and our developed rule sets that can be downloaded, which can also be used for training purposes.
» Download data and rule sets
If you have questions on the procedures, or ideas for collaboration, please contact Dr. Norman Kerle
Events the group is contributing to
