DIFFERENTIAL OBJECT EXTRACTION METHODS FOR AUTOMATED GIS UPDATES

In modern geospatial applications object extraction becomes increasingly part of larger cycles of GIS updates. In such update cycles, the objective is to compare new information to older one, and to identify changes that occurred in the meantime. In this paper we present an image-based GIS updating framework and corresponding image analysis algorithms developed by our group to automate GIS updates. More specifically, we present an overview of our work on two different types of objects to be extracted and monitored, roads and buildings, and the corresponding algorithms, namely differential snakes and differential template matching. Our approach to both problems is characterized by the comparison of a new image to pre-existing information through an automated image analysis tool. This is equivalent to comparing an object as it is represented in an image to the same object represented in a GIS at a prior instance, setting up a novel matching problem, whereby an object is compared to itself to identify how it has changed. For both roads and buildings our algorithms make use of accuracy estimates accompanying the pre-existing information, to ensure the meaningful update of geospatial databases. Making use of these accuracy measures we differentiate change detection and versioning. In this paper we present an overview of theoretical issues behind our algorithms and experimental results from the developed software solutions.