Study on Image Segment Based Land Use Classification and Mapping

High-resolution remote sensing images with a clear spatial and texture information are increasingly used for land use information extraction. Traditional pixel-based remote sensing classification methods are unable to meet the high-precision requirements of land use classification. GIS technology has brought innovations in computer graphics technology. Digital land use maps are widely used in land departments. This paper presents an image segment based land use classification and mapping method. First, multi-resolution segmentation is used to create image segments of different class levels. Then, according to the image segments´ features and relationship between them, the fuzzy classification based on knowledge rules is carried out. Subsequently, the classification results are imported to a land use mapping system based on ArcGIS for editing and modification. Finally, cartographic generalization is executed based on knowledge rules and an integrated land use spatial database is built. The experimental area is selected from a scene of multispectral SPOT-4 imagery. The results have a high overall accuracy. The thematic map can conveniently display the status of land use visually and provide decision support for the land departments.