A new approach for the bundle adjustment problem with fixed constraints in stereo vision

A new approach for the bundle adjustment problem with fixed constraints in stereo vision is described in this paper. Since the direct application of traditional bundle adjustment fails to use the inner constraints completely which are maintained by fixating the orientation and the baseline between the left and right cameras. However, if the fixed constraints are applied to the traditional bundle adjustment, we refine only the left camera extrinsic parameters and 3D points for simplification in stereo pairs. The new method using the fixed constraints has superior theoretical 3D accuracy, and it can reduce the matrix dimension of the covariance matrix so that the total computation time is decreased. Experiments results using synthetic and real data have shown that our method is better than the traditional bundle adjustment algorithm in the 3D accuracy and the convergence rate.

Figure optionsDownload as PowerPoint slide