Pose estimation has been widely used in many robotic and computer vision applications, for examples, multi-camera rig calibration and navigation. Some of relative works employed Kalman Filter and its variants for estimating SE(3) elements. and some other works use the Bingham filter and its variants for improving the estimation accuracy. However, the direct use of Kalman filter will cause the continuity of quaternions in manifold space to be destroyed, and the Bingham distribution based methods will increase the computational complexity. Therefore, we develop a new filter which uses a quaternion metric to make an approximation to simplify the calculation while preserving the continuous structure of the quaternion in the manifold space. Then, the new filter is combined with the adaptive filtering to achieve higher accuracy and free users from manual tuning.
あとは、自由に HTML でつくってもらって結構です。なお、全体の分量として は、このページをプリントアウトした時に、A4 一枚程度になるようにしてく ださい。