Differential Snakes for Change Detection in Road Segments

from digital aerial and satellite imagery may be found in Abstract Suetens et al. (1992), Gruen et al. (1995b), Gruen et al. (1997), The automation of object extraction from digital imagery has and Lukes (1998). been a key research issue in digital photogrammetry and comThe majority of current object extraction methodologies puter vision. In the spatiotemporal context of modern GIS, with are semi-automatic, whereby a human operator provides manuconstantly changing environments and periodic database really some approximations (e.g., by selecting points on a monivisions, change detection is becoming increasingly important. tor display) and an automated algorithm uses these points as In this paper, we present a novel approach for the integration input to extract a complete object outline. Considering roads of object extraction and image-based geospatial change deand similar linear features, these approximations may be in the tection. We extend the model of deformable contour models form of an initial point and an approximate direction. This (snakes) to function in a differential mode, and introduce a information is used as input to automated algorithms that pronew framework to differentiate change detection from the ceed by profile matching (Vosselman and de Knecht, 1995), recording of numerous slightly different versions of objects that edge analysis (Nevatia and Babu, 1980), or even combinations may remain unchanged. We assume the existence of prior of both (McKeown and Denlinger, 1988). Alternatively, the information for an object (an older record of its shape available human operator may provide a set of points that roughly in a GIS) with accompanying accuracy estimates. This inforapproximate the road from start to end, e.g., a polygonic mation becomes input for our “differential snakes” approach. approximation of a long road segment. This information is In a departure from standard techniques, the objective of our used by automated methods like dynamic programming and object extraction is not to extract yet another version of an deformable contour models, i.e., snakes (Gruen and Li, 1997; object from the new image, but instead to update the preLi, 1997). Full automation is pursued by automating the selecexisting GIS information (shape and corresponding accuracy). tion of the above-mentioned necessary initial information By incorporating accuracy information in our technique, we (e.g., node locations, road orientation). Examples of substantial identify local or global changes to this prior information, and efforts towards full automation may be found in Baumgartner et update the GIS database accordingly. This process is compleal. (1999) and Barzohar et al. (1997). In Tönjes and Growe mented by versioning, where, in the absence of change, the (1998), the authors use a knowledge-based approach (semantic pre-existing information may be improved in terms of accunets) to combine information from multiple sensors for road racy if the new image so permits. Experimental results (using extraction. synthetic and real images) are presented to demonstrate the Research on image-based geospatial change detection is performance of our approach. rather limited, at least compared to the body of work on object