Spatial Accuracy Assessment of Object Boundaries for Object-based Image Analysis

Before an object-based image classification can be used, one has to evaluate if it qualifies for its designated purpose, because uncertainty occurs in the complex process of object-based image analysis (OBIA). The range of potential OBIA applications produces objects of a variety of conceptualisations that require a flexible validation concept based on the notion of usefulness for an intended purpose. Objects of a relatively simple notion can be judged into ‘right’ or ‘wrong’, but validation concepts need to be extended to a spatial accuracy assessment of the objects’ boundaries. This paper applied an approach of object comparison, Object Fate Analysis (OFA), for investigating the deviation between boundaries of an OBIA land use/land cover classification and a reference dataset fused from multiple image interpretations. With OFA different error band widths were analysed to determine the amount of deviation of the boundaries. Of the OBIA derived objects 75% extruded less than 6.5 m over the boundary of their corresponding reference object. * Corresponding author.