Sushi: Supervisory Scanner Human Interface
Michael Greenspan

Object digitization - the process of acquiring a complete covering set of range images of the surface of a model - is a highly skilled process, and can be labor intensive.  Sushi is a system which applies a semi-autonomous approach to the human computer interaction of digitization.
 

The Sushi system was inspired by previous research in the supervisory control of robots, as well as the Hymarc Ltd. Rapidscan^TM project. The hardware platform comprises a Vitana Biris range sensor mounted on a CRS A465 robotic manipulator. The operator is provided which a 3D simulation visualization of the  robot-mounted scanning system and workspace, as well as a rich set of tools to help plan and visualize image acquisition.
 

Current system features include:

• frustum positioning and gaze planning tools
• collision detection/motion planning
• ICP-based registration
• image filtering and post processing

An attractive aspect of the Sushi framework is that it is incremental: additional functions with a greater degree of autonomy can be added as they become available.

Following are some images of two objects that we have modeled using Sushi - a Brachiosaurus and Mr. Clown (thank you Isaac for lending us some toys). Each model required around 20 range images. It took approximately 4 hours to acquire the data which, given the complexity of the objects, is quite efficient.

Brachiosaurus

Mr. Clown