Volumetric Reconstruction of Models with Realistic Texture Mapping

The goal of this research is to construct highly detailed 3D thermal models of tires using multiple sets of thermal images generated from simulated heated tires. For complex inspection systems involving heavy equipment, the cost of design is a critical issue. Additionally, system performance must be proven before the purchase of expensive equipment and sensors. Simulation provides a relatively easy and inexpensive means for modification of both the inspection system and the algorithms for producing high detailed 3D models. For instance, the camera placement can be optimized off-line before construction of the prototype. Additionally, the algorithms can be studied and optimized to meet both current and future needs of the tire inspection system. As proof of concept, current research targets the mapping of thermal images onto the surface of a tire. The benefits of this system are that it intrinsically ties a point on the 3D model to the thermal image, something not possible with 2D images alone. This system can also be used with many different models of tires; the operator does not need an accurate model of the tire, it will be reconstructed.

Methodology and Results:
The initial phase of this research involves taking multiple range images of the tire. These range images can be driven by a next best view system, or the user can specify the positions manually. Thermal images are then taken at the same viewpoints. At this time the thermal data can be used as a texture map on the surface mesh generated from the range data. These multiple views of the object are 2^1/2D. The next step is to get a full 3D reconstruction of the object. This goal can be accomplished by a volumetric reconstruction technique. A volumetric approach consists of building a discretized workspace composed of volumetric cells. Each physical cell is associated first with an occupancy status. The occupancy status indicates whether the simulated range sensor captured (x,y,z) information about the tire under inspection. Moreover, a cell is associated with texture coordinates referring to the thermal information captured from the simulated IR camera. Each view is integrated into the discretized workspace and a marching cubes algorithm is used to create a triangle mesh representation of the gathered data. The integration method produces a triangle-meshed, 3D model textured with multiple 2D thermal images.

This work is being conducted at IRIS lab by W. Harvey Gray under the supervision of C. Dumont, and M. A. Abidi (Thesis Chair). This work is supported by RADIAN Inc.'s research program under grant B01998048.

• Papers

H. Gray, E. Juarez,  C. Dumont and M. A. Abidi,  Simulation of a tire inspection system, QCAV'99, Les trois-rivieres, Quebec (Canada), May 18-21, 1999.

L. D. Han, M. Qureshi, M. A. Abidi, H. Gray, Truck rollover warning system simulations, accepted to ISATA, Vienna, Austria, 14-18 June 99