29 number of framework was set expertly. The next step affected by author of classification is the choice of training objects. For each of the classes 5 objects were considered that cover the range of variability of natural conditions. However, their selection was also a affected by author's choice. Verification and comparison of the results of object-oriented analysis with traditional methods of pixel-based classification showed differences in the of classification methods. In the case of object- oriented analysis spectral information are evaluated for the whole object. The highly heterogeneous objects such information may be averaged out and the object is thus classified as atypical representative of a certain class. Pixel classification methods, however, assess the quality of each pixel of the raster, so that any heterogeneous objects are divided into several classes. Using the results of the supervised classification the accuracy of object-oriented classification has been confirmed, considering a few exceptions. 4.1 THEORETICAL ISSUES Serious question of developing comprehensive landscape typology is the purpose and reason for such action. Pedroli et al. (2006) points to a general demand for accurate, highly detailed and high quality representative spatial information on the status and development of...
| Identifer | oai:union.ndltd.org:nusl.cz/oai:invenio.nusl.cz:278399 |
| Date | January 2010 |
| Creators | Romportl, Dušan |
| Contributors | Lipský, Zdeněk, Kolejka, Jaromír, Hrnčiarová, Tatiana |
| Source Sets | Czech ETDs |
| Language | Czech |
| Detected Language | English |
| Type | info:eu-repo/semantics/doctoralThesis |
| Rights | info:eu-repo/semantics/restrictedAccess |
Page generated in 0.0014 seconds