The use of landscape pattern metrics in remote sensing image classification

Abstract

This letter demonstrates the utility of landscape pattern metrics for increasing classification accuracy of land cover. Three examples are provided for the use of a landscape shape complexity measure in remote sensing segmentation‐based classification schemes. The examples include: (1) classification of thaw lakes on the North Slope of Alaska; (2) classification of drained basins in Alaska; and (3) classification of natural vs. anthropogenic pastures in Bolivia. In these examples the Square Pixel Metric (SqP) was applied to objects created from image segmentation to distinguish between categories that had similar spectral properties but different shape complexity values.

Keywords:

• Landscape metrics
• Shape complexity
• Spatial classification
• Segmentation

Acknowledgments

Image processing was performed with ENVI and IDL 4.0 software produced by Research Systems, Inc. and eCognition 4.0 software produced by Definiens Imaging. The author is grateful to Molly Bantz of Research Systems, Inc. and Markus Heynen of Definiens Imaging.