Automated Land-use Map Production from Aerial Photographs of Puerto Rico Using a Contextual Reclassification Algorithm

The Xplorah Planning Support Systems has been under development since 2000 to provide the Puerto Rico Planning Board with an advanced capacity to forecast land-use changes as the result of various scenarios and to assess alternative planning and policy options in their fully integrated, dynamic and spatial context. To fully deploy the analytic capabilities of Xplorah, high quality geographically referenced data are a necessity. These are typically available from Geographical Information Systems at the Planning Board, or other government agencies. The most crucial of all GIS data layers are high quality time series of land-use maps. In order to facilitate regular updating of the land-use map featuring all 21 land-use classes of Xplorah, while assuring a high accuracy, an automatic image processing procedure is being developed. In this procedure the contextual reclassification algorithm OSPARK (Optimised Spatial Reclassification Kernel) is applied to level-sliced, segmented pre-classification of digital aerial photographs with blue, green, red and near-infrared bands. Like SPARK, it is a kernel-based reclassification algorithm, however it automatically adapts the kernel size as a function of spatial variation in the neighbourhood of each pixel to be classified. The algorithm uses the configuration of objects within a kernel in order to effectively discriminate functional land-use classes and land cover. This paper presents the automatic classification procedure and evaluates the results using independent land-use data derived from visual interpretation. Suggestions to improve the procedure are given.