Fast and Accurate Camera Parameter Estimation Based on Feature Landmark Database for Augmented Reality
スポンサーリンク
概要
- 論文の詳細を見る
In the field of augmented reality (AR), a number of vision-based extrinsic camera parameter estimation methods have been proposed to achieve geometric registration between real and virtual worlds. Previously, we proposed a feature landmark-based camera parameter estimation method for outdoor AR applications. Advantages of this method were that a feature landmark database can be automatically constructed by the structure-from-motion (SfM) and there is no necessity to arrange artificial markers in a target place. However, the previous method could not work in real-time because this method involves high computational matching cost between landmarks in a database and image features in an input image. Additionally, the accuracy of estimated camera parameters was insufficient for specific applications which need to overlay CG objects at the position close to the user's viewpoint. This is due to the difficulty in compensation of visual pattern change of close landmarks only from sparse 3-D information obtained by the SfM. In this report, we achieve fast and accurate feature landmark-based camera parameter estimation by employing the following approaches. (1) The number of matching candidates is reduced to achieve fast camera parameter estimation using tentative camera parameter and priority of landmark. (2) Image templates of landmarks are adequately compensated for by considering local 3-D structure of landmark using dense depth information obtained by a laser range sensor. To demonstrate the effectiveness of the proposed method, we have developed several AR applications.
- 2011-05-12
著者
-
Naokazu Yokoya
Nara Institute Of Science And Technologynara Institute Of Science And Technology
-
Tomokazu Sato
Nara Institute Of Science And Technologynara Institute Of Science And Technology
-
Takafumi Taketomi
Nara Institute of Science and TechnologyNara Institute of Science and Technology