The main purpose of the thesis is to perform object-oriented analysis of satellite images, using open source program Orfeo ToolBox. The first part describes object-oriented data analysis and its advantages and disadvantages compared to pixel-based image classification. Following is a description of the steps of object-oriented analysis (segmentation, object classification, post-classification and estimation of accuracy of results). The second part of the thesis describes the open-source program, Orfeo ToolBox, used to implement object classification. The third part describes used data and the area where the classification was performed. The fourth part is a presentation of the processes of the object classification in the selected program. All the steps and procedures for performing segmentation, selection of classes and learning samples, classification, post-classification and accuracy assessment are described.
|
