Progressing from object-based to object-oriented image analysis
This research describes an advanced workflow of an object-based image analysis approach. In comparison to the existing two-staged workflow where typically a segmentation step is followed by a classification step, a new workflow is illustrated where the ob
- PDF / 1,377,952 Bytes
- 14 Pages / 439.37 x 666.142 pts Page_size
- 6 Downloads / 181 Views
M. Baatz 1, C. Hoffmann1, G. Willhauck 1 1
Definiens AG, Munich, Germany; {mbaatz, choffmann, gwillhauck}@definiens.com
KEYWORDS: Segmentation, object-based image analysis, objectoriented image analysis, region-specific segmentation, feature extraction. ABSTRACT: This research describes an advanced workflow of an object-based image analysis approach. In comparison to the existing twostaged workflow where typically a segmentation step is followed by a classification step, a new workflow is illustrated where the objects themselves are altered constantly in order to move from object primitives in an early stage towards objects of interest in a final stage of the analysis. Consequently, this workflow can be called “object-oriented,” due to the fact that the objects are not only used as information carriers but are modelled with the continuous extraction and accumulation of expert knowledge. For better demonstration, an existing study on single tree detection using laser scanning data is exploited to demonstrate the theoretical approach in an authentic environment.
1 Introduction Recent developments in remote sensing made it possible to obtain data of a very high spatial resolution which allows extraction, evaluation, and monitoring of a broad range of possible target features. At the same time, the demand to automate image analysis in operational environments is constantly growing. However, the variety and number of different features to
30
M. Baatz, C. Hoffmann, G. Willhauck
be extracted, add challenges specifically in terms of modelling and autoadaptive procedures. The advantage of a spatial resolution with pixel sizes significantly smaller than the average size of the object of interest comes with the disadvantage of an abundance of spatial detail and the accordingly huge amount of data to be processed. To overcome this drawback, the objectbased image analysis approach has proven to be an alternative to the pixelbased image analysis and a large number of publications suggest that better results can be expected (Baatz and Schäpe 2000, Willhauck et al. 2000, Hay et al. 2005, Kamagata et al. 2005, Manakos et al. 2000, Whiteside et al.2005, Yan et al. 2006). The object-based approach suggests a two-staged approach. In the first step pixels are merged to object clusters, possibly in a multi-level object hierarchy, which then will be analysed and classified in the second step. This means that, the created objects influence the classification result to a large extent although they might not represent the final objects of interest (i.e. single buildings, trees, etc.) already. Because the objects remain unchanged once they are created, and subsequently serve as basis for the actual analysis, this workflow can be called “object-based image analysis”. A successful object-based image analysis results in the correct labelling / classification of regions rather than extracting final objects of interest for instance like trees, acres, buildings or roads in their final shape. In comparison to the “object-based” workflow, thi
Data Loading...
