ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., II-3/W1, 27-33, 2013
© Author(s) 2013. This work is distributed
under the Creative Commons Attribution 3.0 License.
16 May 2013
GeoICT lab, Department of Earth and Space Science and Engineering, York University, 4700 Keele Street, Toronto, ON M3J 1P3, Canada
Keywords: Airborne LiDAR, Airborne Imagery, Rooftop Modelling, Model Selection, Sequential Modelling Abstract. This paper presents a sequential rooftop modelling method to refine initial rooftop models derived from airborne LiDAR data by integrating it with linear cues retrieved from single imagery. A cue integration between two datasets is facilitated by creating new topological features connecting between the initial model and image lines, with which new model hypotheses (variances to the initial model) are produced. We adopt Minimum Description Length (MDL) principle for competing the model candidates and selecting the optimal model by considering the balanced trade-off between the model closeness and the model complexity. Our preliminary results, combined with the Vaihingen data provided by ISPRS WGIII/4 demonstrate the image-driven modelling cues can compensate the limitations posed by LiDAR data in rooftop modelling.
Received: 27 Mar 2013 – Revised: 07 May 2013 – Accepted: 10 May 2013 – Published: 16 May 2013