Pose Estimation for Augmented Reality Applications Using Genetic Algorithm
By Ying Kin Yu, Kin Hong Wong and Michael Ming Yuen Chang
Department of Computer
Science and Engineering
The Chinese University of Hong
Kong
Last updated on 20th December, 2005.
Abstract
This page describes a genetic algorithm that tackles the pose estimation problem in the field of computer vision. Our genetic algorithm can find the rotation and translation of an object accurately when the 3D structure of the object is given. In our implementation, each chromosome encodes both the pose and the indexes to the selected point features of the object. Instead of only searching for the pose of the object as in many of the existing work, our algorithm at the same time searches for a set containing the most reliable feature points in the process. This mismatch filtering strategy successfully makes the algorithm more robust under the presence of point mismatches and outliers in the images. Our algorithm has been tested with both synthetic and real data with good results. The accuracy of the recovered pose is compared to two existing algorithms. Our approach outperformed the Loweˇ¦s method and another genetic algorithm by Hati and Sengupta under the presence of point mismatches and outliers. In addition, our algorithm has been used to estimate the pose of a real object, which is applicable to augmented reality applications. For example, the pose obtained was used for inserting artificial objects into an augmented reality movie in this page.
Quick Links
ˇ@
In this experiment, the test image sequence was taken by putting a paper
box on a rotating turntable. Images were captured using a commercial web camera at a constant time interval. The KLT tracker was used to extract feature points and track them in the image sequence. The known structure, together with point features, were used by our genetic algorithm to track its pose in the image sequence. After the acquisition of the object's pose, three feature points on the paper box in the first image were selected to define a plane for the placement of the virtual object. The synthetic chair was then placed on this plane using the pose sequence computed by our algorithm. Below is the demonstration video that shows the effects of augmenting the synthetic chair (in blue color) with the real scene.ˇ@
Demonstration video (posega_demo_revised.mpg)
Ying Kin Yu, Kin Hong Wong and Michael Ming Yuen Chang, ˇ§Pose Estimation for Augmented Reality Applications Using Genetic Algorithmˇ¨, IEEE Transactions on Systems, Man and Cybernetics-Part B, vol.35, no. 6, pp. 1295-1301, December 2005.
Ying Kin Yu, Kin Hong Wong and Michael Ming Yuen Chang, ˇ§An Evolutionary Approach to the Pose Estimation Problem of Mixed Realityˇ¨, submitted to a conference.
Contact Us
Email: ykyu
ˇ@