Location based service and AR(Augmented Reality) techniques have been one of the most important areas in current computer games and mobile application. This paper presents GPU(Graphics Process Unit) based image recognition system for outdoor games and mobile application. In learning stage, we extract features from input image sequence and convert them into meta information in database. User captures then a query image that is sent to GPU based server system. The proposed system build a descriptor codebook using BRISK and SURF descriptors, and employs codebook with re-ranking information based on homography.

ABSTRACT
 1. 서론
 2. 특징추출 및 기술자 생성
  2.1 SURF
  2.2 BRISK
  2.3 성능 비교
 3. 위치 인식 시스템
  3.1 계층적 k-means 트리기반 FLANN 구성
  3.2 기술자 코드북 생성
  3.3 GPGPU 기반 알고리즘 구현
 4. 실험결과
 5. 결론
 감사의 글
 참고문헌

• 강성일(Dept. of Imaging Science and Arts, GSAIM, Chung-Ang University) | Sungil Kang
• 홍현기(Dept. of Imaging Science and Arts, GSAIM, Chung-Ang University) | Hyunki Hong Correspondence to