Parameter Selection for Segmentation in Object-Oriented Classification of Remotely Sensed Imagery
Shukui Bo, Xinchao Han · 2010
In object-oriented classification of remote sensing imagery, image segmentation is the first step and its quality has significant effect on resulting classification. The quality of image segmentation is always controlled by user-supplied parameters. However, there is not a common way to guide the user selecting a suitable parameter for image segmentation. This paper focuses on the problem of parameter selection for region-growing method, which is one of the most popular segmentation techniques in object-oriented classification of remotely sensed imagery. The presented method selects the suitable parameters by means of training sample areas of each class chosen from an image. The parameter selection method is verified in an experiment of object-oriented classification.