3-D urban modeling from satellite images is of recent interest because it is now easy to obtain web-based satellite images. This paper proposes a new method for 3-D urban modeling using satellite images and building blueprints. Firstly, we derive each building’s structure from its blueprint. Secondly, we automatically estimate the relative pose of the building from satellite images using an MCMC (Monte Carlo Markov Chain). The mean shift segmentation algorithm and the Canny edge detection method are used for the MCMC cost function. We use the shadow of the building to construct a jumping distribution function for the MCMC. Finally, we evaluate our algorithm by comparing scenes from the air with building scenes synthesized by our algorithm.
|저 자||Kil-Ho Son, Jihwan Woo, In So Kweon|
|학 회||The 13th International Conference on Advanced Robotics|