An intelligent vector agent processing unit for geographic object-based image analysis

The consideration of meaningful geographical objects instead of individual pixels is the main distinguishing characteristic of Geographic Object-Based Image Analysis (GEOBIA) methods. In this regard, meaningful geographical objects enable image analysis methods to consider extra information, such as geometry, texture and spatial relationships to analyse imagery. However, meaningful geographical objects are determined with only limited spatial knowledge of real world objects. Furthermore, such objects remain associated with a static geometry that depend greatly on user-defined parameters such as scale and colour during segmentation. This paper explores Geographical Vector Agents (GVAs) as an automated processing unit to address the above-mentioned limitations of the image objects, allowing them to evolve in terms of geometry and attributes during a dynamic and concurrent segmentation/classification process. Experimental tests indicate satisfactory results of the capabilities of GVA s for generic applicability in GEOBIA methods.