영상기반 헤드트랙커 시스템에서 기하학적 해싱을 이용한 기하학적 패턴인식

Abstract

This paper presents the optical head tracker system that estimates the position and attitude of the helmet from the stereo image of two CCD cameras. This head tracker consists of two infrared CCD cameras, infrared LEDs that are stuck on the helmet and computer for image processing. LEDs are a feature points that is arranged in a specific pattern. The algorithm of this system consists of feature segment process, projective reconstruction process and pattern reconstruction process. Pattern reconstruction process that is described in this paper classifies into the same feature points in each frame to obtain the position and attitude of the helmet. In this process, it uses geometric hashing which is a representative technique for model based object recognition in computer vision. But, the performance of geometric hashing degrades rapidly as the noise increase. So, this paper describes a geometric pattern recognition using a modified Haussdorf distance in voting process unlike the general method of geometric hashing.