Multi-scale Landscape Pattern Analysis Based on High Spatial Resolution Imagery Segmentation

The development of high spatial resolution remote sensing, such as IKONOS and QuickBird, makes it possible to describe the landscape pattern in the smaller scales. However, the level of landscape pattern detail represented by high spatial resolution images is different from middle and low resolution images. The image process and image analysis methods are also different as well. New techniques including image segmentation and object-based analysis are widely used in the high spatial resolution imagery analysis. So the landscape models and analysis methods set up by middle and low spatial resolution remote sensing are not suitable for the high spatial resolution remote sensing. In this study, focusing on the method for landscape pattern analysis for high spatial resolution remote sensing, we try to achieve the following goals: 1) set up a hybrid landscape model to describe the landscape patch mosaic, continuum and connectivity characteristics based on a unified data structure; 2) set up a landscape scaling method based on high spatial resolution image object-oriented analysis. The result shows that Delaunay-Voronoi data structure based on image object-oriented analysis is proper to set up the hybrid landscape model. Multi-scale image segmentation is better than aggregation method for multi-scale landscape pattern analysis.