Development of a land use extraction expert system through morphological and spatial arrangement analysis

• A residential and commercial land use (LU) extraction system is developed.
• We develop a post-classification correction model using spatial arrangement analysis.
• The model accuracy improves through post-classification correction process.
• The two LU classes are extracted with a very high accuracy.
• The proposed model divides zones with mixed LUs into subzones with single-LU types.

Land use (LU) information is of significant value for various urban studies and is needed for a wide variety of decision-making initiatives in the range of global, regional and urban areas. A challenge that researchers and practitioners have been facing in urban modeling/planning is the lack of detailed information regarding how cities are structured and how urban development evolves. This study aims to develop a hierarchical rule-based LU extraction framework using geographic vector and remotely sensed (RS) data, in order to extract detailed subzonal LU information. The LU extraction system, which considers both morphological and spatial arrangement analyses at a fine spatial level – parcel, is developed to understand association/correlation rules between different urban features and their corresponding LU structures. In this study, structures and patterns of residential and commercial LUs are scrutinized. Residential and commercial LUs are first extracted by examining the morphological properties of individual parcels using a stepwise binary logistic models, which results in an overall accuracy of 97.5% and 92.4% respectively. A spatial arrangement analysis is then carried out through Gabriel Graph to identify structural patterns of residential and commercial parcels in order to cluster and separate them from other LUs. Extracting residential and commercial clusters helps to correct misclassifications arising from morphological analysis. The post-correction process results in improving the overall LU extraction accuracy by 1.6% for residential and 4.8% for commercial LU. The above exercises show that the LU classification framework developed can classify and then divide large zones with mixed LUs into single-LU subzones with a high accuracy.