Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
An improved augmented reality system based on AndAR
Chen P., Peng Z., Li D., Yang L. Journal of Visual Communication and Image Representation37 (C):63-69,2016.Type:Article
Date Reviewed: Aug 23 2016

Chen et al. introduce an augmented reality (AR) mechanism in mobile computing using AndAR as the AR development kit for developing augmented reality applications in Android smartphones. This framework has 3D registration based on markers; however, registration based on natural features is much more practical. The AndAR framework is open source, so it may be easily modified to include new functionality. Therefore, the authors adapt this framework, making it capable of registration based on both markers and natural features.

The process has several steps: obtain a frame, detect the target (either the marker or a set of natural features that determines the target object) and the camera extrinsic matrix, and render the new image with the virtual object using the camera extrinsic matrix. This process is described clearly and step by step. In addition, the required adaptations in the framework are highlighted.

For registration using natural features, at least four pairs of correct matching points (features) are required. Registration is divided into two stages: feature detection and descriptor matching. The feature detection method selected in this paper is ORB, which is a comparison method between the target object and the captured frame. The set of ORB descriptors are evaluated using Hamming distance to measure the degree of similarity between them. With these matches, the homography matrix is computed. In order to deal with each different frame, which could provide spatial jittering of features, the Lucas-Kanade optical flow method is applied. Using this mechanism, a reduction in tracking feature points is obtained.

The proposed system is evaluated using just one mid-range smartphone (Huawei Honor 3C) and just one experiment. It produces good visual results with two virtual objects. Registration using natural features with ORB descriptors provides a frame rate of just 2.2 fps, whereas the one using ORB descriptors and the Lucas-Kanade optical flow filtering method yields 12.7 fps. The authors conclude that this system meets real-time requirements. However, in my opinion, at least 24 fps (or some figure around this value) is required to state that real time is achieved.

This paper is highly technical. The authors describe each modification in the code to make the AndAR framework manage registration with natural features. The experimental setup is could clearly be improved, as just one experiment is described. The paper is recommended reading for developers in the initial stages of creating augmented reality applications for smartphones.

Reviewer:  José Manuel Palomares Review #: CR144707 (1611-0841)
Bookmark and Share
  Featured Reviewer  
Artificial, Augmented, And Virtual Realities (H.5.1 ... )
Registration (I.4.3 ... )
Programming Environments (D.2.6 )
Processor Architectures (C.1 )
Would you recommend this review?
Other reviews under "Artificial, Augmented, And Virtual Realities": Date
Benedikt M. (ed), MIT Press, Cambridge, MA, 1991. Type: Book (9780262023276)
Dec 1 1992
Knowledge-based augmented reality
Feiner S. (ed), MacIntyre B., Seligmann D. Communications of the ACM 36(7): 53-62, 1993. Type: Article
Aug 1 1994
Playing God
Roehl B., Waite Group Press, Corte Madera, CA, 1994. Type: Book (9781878739629)
Mar 1 1995

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2023 ThinkLoud®
Terms of Use
| Privacy Policy