CVSL Logo
FrancaisHome
AboutPeopleResearchPublicationsEventsProfile
About
REPARTI Seminars


The REPARTI Seminars at Université Laval are held on Fridays at 11:30 a.m.
Please see the program for more details.
Dec 14 2017 1:30PM
Seminar
Deep 6-DOF Tracking
Dec 15 2017 11:00AM
Seminar
Building and Evaluating Data-Driven Neural Dialogue Systems

 

 

 

REPARTI

MIVIM

Aug 17 2007 11:00AM

Flavio Augusto Prieto Ortiz, professeur associé, Departamento de Ingeniería Eléctrica, Electrónica y Computación, Universidad Nacional de Colombia Sede Manizales

Real–Time Robot Manipulation Using Mouth Gestures in Facial Video Sequences

Automated Reverse Engineering of Free Formed Objects Using Morse Theory



Abstract

Real–Time Robot Manipulation Using Mouth Gestures in Facial Video Sequences

A novel method for the automatic command of three degrees of freedom of a robot using mouth gestures is presented. The method uses a set of different pixel–based segmentation algorithms and morphological restrictions in order to extract the mouth area from the frames. A fuzzy inference system is then used in order to produce a small subset of discriminant features for gesture classification. A state machine was designed in order to stabilize the robot command task by using a temporal sliding mean on the detected gestures. Experimental results show that the method is both robust and reliable when operated by different people, and fast enough to keep the detection’s rate in real–time.

Automated Reverse Engineering of Free Formed Objects Using Morse Theory

A method for surface reconstruction by means of optimized NURBS (Non-Uniform Rational B-Splines) patches from complex quadrilateral bases on triangulated surfaces of arbitrary topology is proposed. To decompose the triangulated surface into quadrilateral patches, Morse theory and spectral mesh analysis are used. The quadrilateral regions obtained from this analysis are then regularized by computing the geodesic curves between each corner of the quadrilateral regions. These geodesics are then fitted by B-splines curves creating a quadrilateral network on which a NURBS surface is fitted. The NURBS surfaces are then optimized using evolutive strategies to guaranty the best fit as well as C1 continuity between the patches.




     
   
   

©2002-. Computer Vision and Systems Laboratory. All rights reserved