Shape-Based Descriptions to Classify Na Dun Cultural Settlement using Object-Based Image Analysis

Abstract

Rural village forms are classified into 3 main types: Nucleated, Dispersed and Linear villages. The purpose of this paper is to describe rural village shapes using object–based image analysis. This method is demonstrated in the Na Dun district, northeast Thailand. Village form, or village morphology, is a pattern of village components that form an irregular shape via the merging of candidate classes, e.g., roofs, vegetation area, public area, etc. Prior to image object fusion to shapes, village components are classified by rule-based methods with an overall accuracy of 87.33%, and 0.48 of kappa statistics. To do this, an image object fusion algorithm is used to merge the village’s candidate classes (based on spatial constraints) into a shape that represents the village’s form. The shape of each village type is defined by the essential parameters obtained from the component relationships and spatial factors. These factors are derived from pan sharpened QuickBird images with spatial resolutions of 0.6 meters. The study determines shapes by applying geometry shape features, i.e., elliptic fit, compactness, roundness and shape index, by using statistical analysis and customized relational features. These factors are compared to all neighboring objects across the entire image scene on both the vertical and horizontal image object levels to create class descriptions for village from classification.