Pose estimation for multi-camera systems

We build a multi-camera platform and present a relevant approach for pose estimation of the multi-camera system with known internal camera parameters. The proposed approach is able to solve the pose estimation, for a three-camera system that main camera has overlapping views with the left camera and the right camera respectively. By taking advantage of the optical flow method and the key frame policy to track the feature points, using Kalman filter (KF) to manage 3D points and adopting bundle adjustment (BA) to optimize the pose of the main camera, the experiment shows that the pose estimation results of the multi-camera system have good robustness.