Dong-jin HAN, Hern-soo HAHN
(Intelligent Robot Research Center, Soongsil University, Seoul 156-743, Korea)
Abstract-We propose a methodology, which selects informative images from long video sequence. Video camera usually collects many frames, maybe too many. Some of images could have been blurred or cluttered due to the object motion since the camera is mounted on moving platform. For SLAM (simultaneous localization and mapping) using video camera, SfM (structure-from-motion) or any other computer vision task, it may be not efficient or appropriate to use all of the image frames. Several criteria are suggested.
Key words-3D; robot; navigation; feature tracking
Manuscript Number: 1674-8042(2011)supp1.-0019-04
doi: 10.3969/j.issn.1674-8042.2011.supp1.005
References
[1]Lowe D G. Distinctive image features from scale-invariant keypoints. IJCV, 2004, 60(2): 91-110.
[2]Eade E, Drummond T. Scalable monocular SLAM, Proceedings of IEEE Computer Conference on Computer Vision and Pattern Recognition, 2006: 469-476.
[3]Hartley R I, Zisserman A. Multiple View Geometry in computer vision. Cambridge University Press, UK, 2000.
[4]Hartley R, Triggs B, McLauchlan P, et al. Bundle adjustment—a modern synthesis. Vision algorithms: Theory and Practice, 2000, 1883:153-177.
[5]Wang L, Zhao G, Rajpoot N, et al. Special issue on new advances in video-based gait analysis and applications: challenges and solutions. IEEE Trans. on Systems, Man, and Cybernetics, 2010, 40(4): 982-985.
[6]Lemaire T, Berger C, Jung, I K, et al. Vision-based slam: Stereo and monocular approaches. International Journal of Computer Vision, 2007, 74(3): 343-364.
[full text view]
