EXPLORATION OF LANDUSE TYPE IMPORTANCE FOR LANDUSE DATA GENERALIZATION

摘要

Landuse data generalization is one of the significant methods to produce multi-scale and multi-theme data and derive multi-resolution databases from a comprehensive landuse database. Landuse data generalization requires high flexibility and practicability for the generation of multi-scale and multi-theme landuse maps in various application fields. For meeting the special requirements, generalization constraints should vary with the real dataset and so do the corresponding processes for landuse data generalization. This paper focuses on exploring landuse characteristics existing in real datasets to construct generalization constraints. The Shannon Diversity index is used to investigate the distribution of landuse area in order to figure out the possibility in existence of some dominant landuse types. Then, we employ multi-attribute decision model (MADM) to judge further the dominance of each landuse type. With the dominance of landuse types, we present a mathematic model to determine the minimum area threshold for each landuse type. The minimum area threshold is used to construct a basic generalization constraint, that is, all of patches should be larger than the minimum area threshold of the corresponding landuse types. The generalization of a real dataset demonstrates the function of the presented methods and the final output is evaluated and compared with the traditional methods. The experimental results show that the exploratory methods are effective to generalize landuse data and preserve the significant characteristics in a target map as well.