In this paper, we present a method to increase the robustness and the efficiency of template-based object matching. Instead of preparing templates of various scales and deformations, we use only one template and transform the appearance of the target object by placing a virtual camera seeing a very similar appearance to the template. This approach can solve the chronic problem of scale-dependence and distortion in the conventional template matching, and therefore be applied to the object detection and alignment tasks of general camera configuration systems. The proposed method has shown a good performance in the application of the service robot, Vistro, which attended the Robot Grand Challenge 2009 in Pohang, Korea.
조회 수 2407 댓글 0
|저 자||Jiyoung Jung, Yekeun Jeong, Joon-Young Lee, In So Kweon|
|학 회||16th Korea-Japan Joint Workshop on Frontiers of Computer Vision (FCV)|
Prev Dynamic Programming and Skyline Extraction in Catadioptric Infrared Images Dynamic Programming and Skyline Extraction in Catadioptric Infrared Images 2010.01.18by TA Detecting Small Objects in Natural Scene using Depth Cue Next Detecting Small Objects in Natural Scene using Depth Cue 2010.01.14by