Granular Approach to Object-Oriented Remote Sensing Image Classification

摘要

This paper presents a summary of our recent research in the granular approach of multi-scale analysis methods for object-oriented remote sensing image classification. The promoted granular Hough Transform strengthens its ability of recognize lines with different width and length in remote sensing image, while the proposed granular watershed algorithm performs much more coherently with human visual characteristic in the segmentation. Rough Set is introduced into the remote sensing image classification, involving in the procedures of feature selection, classification rule mining and uncertainty assessment. Hence, granular computing runs through the complete remote sensing image classification and promotes an innovative granular approach.