The Scale-Span Classification Research for Multispectral Images Based on the Homogeneous-Region

Existing classification methods which are based on the homogeneous-region mostly involve the best segmentation scale choice. Using the so-called best segmentation scale to respond the subjective defined objects, it would not be the best way for classification. Therefore we propose a simple classification method with high precision. It is a new kind of multi-scale homogeneous-region model, fully uses the longitudinal information which the homogeneous-region model provides, and adopts the scale-span classification method based on decision tree to improve the accuracy, rather than directly carrying on the best scale choice. The experimental result proves the scale-span method is more accurate than sole scale lassification.