Key Points
We recognize planar objects captured by a camera using edge features extracted from individual regions. This supports an attention-focusing user interface for smartphones and tablets that quickly connects real objects to data on the Internet.
Abstract
Imagine pointing a smartphone or tablet at an object and reaching information about it on the Internet. We developed an interface for this kind of interaction. It needs to recognize flat objects in the camera image quickly and accurately.
Existing methods can take a long time to extract the visual clues needed for recognition. We propose examining individual regions of an image and extracting features from their edges: boundaries where brightness changes sharply. These features describe aspects of the object's shape and pattern while reducing the processing needed to recognize it.
In verification tests using smartphones and tablets, the method achieved 99% recognition accuracy with a response time of 0.5 seconds. The study shows how recognizing real objects can connect them to digital information with a short wait for the user.
Publications (Japanese)
- 大平 英貴, 関根 真弘, 西山 正志,
位置ずれを許容するエッジ特徴による物体識別, 情処研報 CVIM184-15, pp. 1-6, November, 2012. - 大平 英貴, 関根 真弘, 西山 正志,
かざすUIで実世界とオンラインの融合,
映像情報メディア学会 冬季大会 特別企画『キネクト，スマートフォンアプリ，AR 等，体感型アプリの世界』, December, 2012. [最優秀賞]
Short movie
