Laurana Wong Electrical Engineering M.S. 1999

An Algorithm and System for Finding the Next-Best-View in a 3D Object Modeling Task

Research Objectives: The research is aimed at defining sensor poses for the purpose of automated object modeling. The sensor locations are computed from the model of the object being reconstructed. This task involves defining a data structure for information gathering. The next best view from which new information will be captured is computed to reveal a maximum amount of information that was still unknown at the previous stage of the reconstruction. A workspace is defined a priori to encompass any object shape. An objective function allows computation of the next best view at every stage of the reconstruction. Once every cell of the workspace is being inspected, a complete triangle mesh model is created constituting the ultimate result of the Next-Best-View system.

Methodology and Results:
An occupancy grid containing visibility status of the workspace cells is used to gather information about the object being reconstructed. An objective function selects the next best among several potential views. For the purpose of performance estimation, a method for choosing potential views has been developed so that the next best system outputs a new sensor location that maximizes the amount of new information to be captured. Potential views have also been chosen such that they are strongly dependent on the object being reconstructed. This approach allows run time and system performance improvement. The marching cubes algorithm has been used to build triangle mesh models. The Next-Best-View system has been tested on many object shapes. 

This work was conducted by Laurana M. Wong while at IRIS lab under the supervision of C. Dumont and M. A. Abidi (Thesis Chair). This work was supported by DOE's University Research Program in Robotics under grant DOE-DE-FG02-86NE37968, and by RADIAN Inc.'s research program under grant B01998048.

• Papers

L. M. Wong, C. Dumont, and M. A. Abidi, Next Best View System in a 3D Object Modeling Task, Submitted to IEEE Conference on Computational Intelligence in Robotics and Automation, Monterey, California, November 8-9, 1999

L. M. Wong, C. Dumont, and M. A. Abidi, An algorithm and system for finding the next best view in a 3D object modeling task, To be presented to Robotics and Applications IASTED Conference in Santa Barbara, CA USA, October 28-30, 1999

L. M. Wong, C. Dumont, and M. A. Abidi, Determining Optimal Sensor Poses in 3D Object Inspection, Conference on Quality Control By Artificial Vision, Takamatsu (Japan), Nov. 98, ISBN: 4-921073-01-5, pp 371-377

 L. M. Wong, C. Dumont, and M. A. Abidi, An Algorithm for Finding the Next Best View in Object Reconstruction, Photonics EAST, Intelligent System and Advanced Manufacturing 98, SPIE Conference on Sensor Fusion and Decentralized Control in Robotics Systems, Boston, Massachusetts, November 1998, 3523, pp 191-200.