The main intent of this graduation thesis is to present the use of selected open source software for object-oriented high resolution remote sensed image analysis and to design operation instructions. In the first part of the thesis the open source software with the emphasis on geodetic and object-based image analysis software is presented. The second part is dedicated to description of basic principles and steps in object-based image analysis (segmentation, classification, post-classification with classification accuracy assessment) and its advantages against the pixel-based image classification. The third part is dedicated to the presentation of the InterIMAGE program and in the fourth part the operation instructions for executing object-based satellite image classification are given. Two methods of image classification are presented: training samples based classification and decision tree based classification. In the operation instructions all the steps, from data import, segmentation, classification and post-classification to data export and descriptive quality assessment of the results are described.
|
