A “Multisensory 3D modeller” was developed to evaluate different sensor principles and sensor fusion for the generation of photorealistic 3D models. The Modeller can be used hand-guided or automated on a robotic system. Thus, robots should be able to build a photorealistic model of their environment by moving around an end effector equipped with different sensors (a so-called Eye-in-Hand system)
The 3d-Modeller consists of a core module and application specific extensions. The core module, hereafter simply called modeller, comprehends two digital cameras, a miniaturized rotating laser scanner and two line laser modules. Furthermore it carries a small embedded computer to provide internal computing power, and a graphical colour display. The modeller can be connected to a handhold with various input buttons or to a robot via an industrial quick changing adaptor. On the sides markers are adaptable for optical pose measurement.
Due to the modular design the system allows hand-guided operation as well as automated robot applications. Three mechanically identical free of play couplers enable quick and easy exchangeability of all components (see Fig. 1). The Modeller can be manually operated or controlled remotely. The display along with the handhold enables menu-driven control.
The housing of the modeller is not completely sealed for optical performance reasons. For automated applications in dusty environment (mainly on robots) the modeller can be set under light overpressure to avoid intrusion of dust. Therefore an air pressure connector exists at the robot interface.
In contrast to the laser range scanner which represents an independent sensor different system components are combined to realize additionally two concurrent sensors. In overall the sensor devices implemented with the 3d-Modeller are
Laser Range Scanner (LRS)
Stereo Camera Sensor (SCS)
Laser Stripe Profiler (LSP), optionally Dual Cross Hair LSP
which are outlined as follows: