GET-Forschungsseminar Abstracts
Hierarchical best view planning for efficient 3D exploration of indoor environments
Haydar M. Baker , GET Lab
Vortrag: Mi. 30.01.2013, 16:30, Raum P 1.4.17
Zusammenfassung:
This work is concerned with finding the next best view location of a range sensor mounted on mobile robot in indoor environments. To accomplish 3D exploration of indoor environments, more than one scan from more than one location are needed. Hierarchical next best view planning presents an efficient solution for this problem. It concentrates on efficient evaluation of sensing constraints and extracting the partial model of the scene.
The main concept of the work is to reduce the time elapsed in evaluate the feasible view before evaluate the registration constraint. Firstly, the feasible views will be extracted from the scene point cloud, then the range image of the point cloud will be produced. Depending on the range image the border points, which will be thereafter calculated, the occlusion planar patches will be determined. Feasible views beside occlusion planar patches comprise the scene partial model, at which the feasible views will be evaluated verifying certain sensing constraints.
In conclusion N next best views could be extracted and evaluated in reasonable time, enhancing the overall process of finding the next best view.