Comparing to traditional pixel based methods of land cover identification, the object based approach allows to increase the set of discriminant features, including elements related to texture, size, shape, widely understood spatial and geographical context. This extent has important influence on digital classification efficiency and final accuracy. The thematic and positional accuracy of extracted and identified objects strongly depends on segmentation stage, which is very crucial one for further steps.
The paper presents results from different variants of object based land cover classification with the main aim to test the influence of segmentation methods and their parameters on the accuracy of land cover form identification based on VHR satellite image. The influence of selected filtering effects applied on certain stage of image analysis procedure was also tested. The object oriented approach was realized based on Definiens Professional set of tools. The analyses were done for spatially and spectrally complex terrain. Due to the fact that final comparisons of thematic accuracy concern the classified objects the right procedure of accuracy evaluation was developed and applied in the study.
The results show certain positives and limitations of applied approaches and defined variants, and confirm the importance of proper selection of segmentation procedure and adequate parameters.
IOS Press, Inc.
6751 Tepper Drive
Clifton, VA 20124
Tel.: +1 703 830 6300
Fax: +1 703 830 2300 firstname.lastname@example.org
(Corporate matters and books only) IOS Press c/o Accucoms US, Inc.
For North America Sales and Customer Service
West Point Commons
Lansdale PA 19446
Tel.: +1 866 855 8967
Fax: +1 215 660 5042 email@example.com