A Real-Time SFM Method in Augmented Reality

Augmented Reality technology is widely used. In this paper, we proposed an efficient method to obtain the satisfied SFM result in real-time using clustering method. First we use Lucas-Kanade method to calculate the optical flow of adjacent frames. We take advantage of the better movement features to calculate the fundamental matrix and use Sturm’s method to restore the camera intrinsic parameters. We further use SVD decomposition to obtain the extrinsic parameters of adjacent frames. In order to obtain more accurate camera intrinsic and extrinsic parameters, we use clustering method to get the reconstruction result and further to amend the intrinsic and extrinsic parameters. Further experiments proved the validity of the method. The proposed method is a real-time algorithm and can meet the requirements of the study of augmented reality.