Conclusions

This paper has presented a new method for object classification in 3D scans. The scans are automatically acquired by an autonomous mobile robot equipped with the AIS 3D laser range finder. The scanner provides external, light independent, bimodal data that is converted to depth and reflectance images. It is shown that both image types, and especially their combination, are a good choice for object detection and classification. The object is classified with complex classifiers that are arranged in a cascade. The classifiers use Haar-like features and an internal object representation is learned with the Ada Boost techniques. Typical vision problems, e.g, shadows or posters on the wall showing distracting objects [#!MPI!#], are avoided by the use of range images.

Needless to say, some work remains to be done. The detected object will be used as an index to a database of 3D models. The model and the position of the detected object can be used as a start position for an ICP based matching in the range data. Furthermore additional rotated features of $45^\circ$ as proposed by Lienhart and Maydt will be used to improve the classification [#!Lienhart_2002!#]. The overall goal is to use an autonomous mobile robot to build 3D semantic maps that contain temporal and spatial 3D information with descriptions and labels about the environment.