Fingers Image Tracking on Omnidirection Camera by Condensation Algorithm

N. Ikoma, M. Sakata, and M. Doi (Japan)

Keywords

Omnidirection camera, dynamic image, finger image,tracking, CONDENSATION algorithm.

Abstract

Tracking of fingers image on omnidirection camera has been investigated which aims at developing of human friendly interface using fingers' gesture. Where the way of using the omnidirection camera is different from a nor mal usage in which the user holds the camera with his/her fingers and move them. The interface can use the motion of the fingers to establish a communication between the user and computer system in human friendly way. To achieve this, it is necessary for the system to track the fingers in the dynamic image of omnidirection camera. We have em ployed CONDENSATION algorithm for the tracking. The algorithm uses pre-defined shape of finger called template, and estimates the parameters of affine transformation to ad just the template to the image. The estimation is performed with many number of particles where each particle has an realization of the affine parameters. Applying three steps called prediction, observation, and selection, we have up dated particles which approximates a posteriori distribution of the affine parameters given the image sequence up to cur rent time. Real image experiments show the performance of the tracking method.

Important Links:



Go Back