屋外型アノテーションシステムにおける高性能化の検討

摘要

Annotation is one of the major applications of Augmented Reality (AR). The key technology of a successful annotation system is the accurate estimation of the pose and position in the real world. Indoor AR systems often use the active sensing for the purposes, such like magnetic transmitters and receivers. However, the active sensing is not practical at outdoor AR, because it tethers the user a particular area (close to the active sources) and requires an infrastructure. The absence of an active sensing makes it difficult to estimate the pose and position. In this paper, an improved hybrid sensing scheme for the outdoor annotation is proposed, that enables accurate positioning and the pose estimation of the viewer. GPS, rate gyroscope and electronic compass are used for the initial estimation of the pose and position. Then the vision based technique is used for accurate estimations. The boundary edges between the buildings and the sky is chosen as the image derived feature. Matching is done between the image derived boundary edges and the synthesized one from 3D maps. Based on the similarity measure defined by DP matching score, the quasi-optimal estimation is computed using PSO (Particle Swarm Optimization). The evaluation of the proposed method is done through the experiments performed in the campus.%屋外で撮影した建物の情報を景観画像に重畳表示するアノテーションシステムの性能を改善した.GPS，ジャイロ,電子コンパスから得られた観測者の初期位置をカメラから得られた画像特徴量を用いて補正することでアノテーションの際に違和感を与える原因となる幾何学的ずれを解消する.本論文では高精度化と高速化を実現するための手法を物理センサ，ビジョンセンサの両面からそれぞれ提案する.また，従来手法では正確な位置姿勢が一意に推定できなかった特徴に乏しい地点に対する対応についても述べる.