Tomographical object detection using a reduced number of projection angles

Inhalt:
Tomography is in the medical domain used to reconstruct the interior of bodies in a non-descructive way. It allows recognition of 3-D object configurations and can thus be used to detect unknown or modified objects, i.e. independent from shape or other features. Visual light based tomography is thus an interesting approach for detection of unknown or possibly degraded objects. However, for most real world object detection problems only a reduced number of scene views can be provided, which leads to serious challenges for the purpose of a robust object detection. The paper gives a short introduction into visual light based tomography and presents object detection schemes based on a reduced number of images of a scene. Different object detection setups are evaluated and discussed. The resulting configuration requires a set of 5 views that can be acquired by different camera configurations.