IRIS Research Staff

Faysal Boughorbel
Ph.D. Candidate

  Office 410 Science & Engineering
The University of Tennessee
Knoxville, TN 37996-2100
Telephone (865) 974-9213
Fax (865) 974-5459
E-mail fboughor@utk.edu
 

 

Current Research

Modeling Scenes and Objects from Multi-modal Imagery using Gaussian Fields

The focus of my research is on building 3D representations of real world scenes and objects using multiple sensors. Primarily range acquisition devices (such as laser scanners and stereo systems) that allow the recovery of 3D geometry, and multi-spectral image sequences including video and thermal IR images that provide additional surface characteristics. One of the most important technical challenges that I am addressing is the registration task in both its multi-modal and single modality aspects. My work is based on a unified approach that formulates the Correspondence Problem as an optimization task. In this context I developed a criterion that can be used for 3D free-form shape registration, feature-based image registration, and 3D to 2D pose recovery. The new criterion is derived from simple combinatorial matching principles by approximation and relaxation. One of the main advantages of the proposed approach is convexity in the neighborhood of the alignment parameters and continuous differentiability, allowing for the use of standard gradient-based optimization techniques. Physically it is interpreted as an integration over a Gaussian Potential Field. Such formulation proved useful for controlling and increasing the region of convergence, and hence allowing for more autonomy in correspondence tasks. Furthermore the criterion can be computed with  Linear Complexity using recently developed Fast Gauss Transform numerical techniques. The resulting algorithm was applied to several real world problems and proved more effective than current methods.

 

Current Projects
 
 

  

Free-Form 3D Shape Registration using Gaussian Fields - For fully automatic and accurate registration of free-form 3D shapes we devised a force-field inspired technique that overcomes several problems with standard methods such as the Iterative Closest Point (ICP) algorithm. We obtained a smooth convex criterion that can be easily optimized using gradient-based techniques and that extends considerably the range of convergence.   


 

Multi-sensor Shape-based Image Alignment using Gaussian Fields- One of our research interests is the use of Thermal-IR and  Visual images for face recognition and other multi-modal surveillance tasks. The combination of both modalities enhances the performance of our systems. An important pre-processing step is the registration of images provided by IR and Color cameras. To avoid an expensive hardware solution we are developing matching techniques that will be used along with different motion models.   The matching algorithm is based on the Gaussian Fields criterion applied to feature maps extracted from the images. We are also applying the same approach to other multi-sensor correspondence problems


Publications

F. Boughorbel, A. Koschan, B. Abidi, and M. Abidi, "Gaussian Fields: a New Criterion for 3D Rigid Registration," Pattern Recognition, Vol. 37, No. 7, pp. 1567-1571, July 2004.

F. Boughorbel, A. Koschan, B. Abidi, and M. Abidi, "Gaussian Energy Functions for Registration without Correspondences," 17th International Conference on Pattern Recognition (ICPR), Cambridge, UK, August 2004.

F. Boughorbel, B. Abidi, and M. Abidi, "Registration of Infra-Red and Color Images for Multimodal Face Recognition," The Biometric Consortium Conference (BC2004), Crystal City, VA, September 20-22, 2004.

F. Boughorbel, A. Koschan, and M. Abidi, "Multi-sensor Registration and Integration for Inspection and Navigation," 10th International Conference on Robotics and Intelligent Systems for Hazardous Environments, Gainesville, Florida, March 2004.

F. Boughorbel, A. Koschan, and M. Abidi, "Registering Multi-sensor Datasets from a Robotic Inspection Platform," SPIE Defense and Security Symposium, Orlando, Florida, April 2004.

F. Boughorbel, P. Crilly, A. Koschan, and M. Abidi, "Estimating 3D camera motion without correspondences using a search for the best structure," Pattern Recognition Letters, Vol. 24, No. 1-3, pp. 327-337, January 2003.

F. Boughorbel, A. Koschan, and M. Abidi, "Modeling 3D Objects from Range Maps and Color Images using a Warping-based Approach," Proc. 6th International Conference on Quality Control by Artificial Vision QCAV03, Gatlinburg, TN, USA, SPIE Vol. 5132, pp. 288-295, May 2003.

F. Boughorbel, Y. Zhang, S. Kang, U. Chidambaram, B. Abidi, A. Koschan, and M. Abidi, "Laser Ranging and Video Imaging for Bin Picking," Assembly Automation, Vol. 23, pp. 53-59, March 2003.

F. Boughorbel, B. Abidi, S. Kong, and M. Abidi, "Reconstructing 3D Faces from Video, Range, and Thermal Imagery," The Biometric Consortium Conference (BC2003), Crystal City, VA, September 22-24, 2003.

F. Boughorbel, D. Page, C. Dumont, and M. A. Abidi, "Registration and Integration of Multi-Sensor Data for Photo-realistic Scene Reconstruction" Applied Imagery Pattern Recognition '99, SPIE Vol. 3905, pp. 74-84, Washington, D.C., October 13-15 1999.

 

 
Last updated:
Webmaster