立体物の形状と深度認識を利用しＡＲデモ - Depth-assisted Real-time 3D Object Detection for Augmented Reality #AR
形状認識ＡＲでは、単に見かけだけの形ではなく、光や影をどの程度推測して認識できるかがポイントになると感じた。もちろん、目的に応じて精度を求めないことも多い。
In form recognition AR, it was thought that it became the point how much not an apparent form but light and a shadow can only be guessed and recognized.
Of course, according to the purpose, accuracy is not searched for in many cases.
This video demonstrates our work published in ICAT2011 conference. We combine RGB information with 3D depth information for robust target object detection under difficult situations, such as varying lighting conditions, poor textures, and scale differences.
ビジョンの特徴点がどのようになっているかが目で見てわかるデモ動画
The demonstration animation looked at and understood what the details of the vision have become by the eye
We propose a vision-based all-in-one solution for augmented reality where users want to exploit unknown 3D objects in their systems.
In our solution, we facilitate two time-consuming o -line pro-cesses: obtaining information, such as keyframes and keypoints, for real-time tracking of unknown 3D targets, and estimating local coordinates with a scale for accurate registration of virtual content.
The proposed solution only requires images with minimal interactions.
The users do not need to know about 3D marker-less tracking in depth. At the end, we propose a framework for AR miniatures systems to verify the e ectiveness of our solution. As a result, all developed systems worked in real-time, more than 25 framerate, and showed reliable registration even in severe viewpoint changes.
Our demonstration videos are available in the supplemental materials.
