イベントベースカメラを用いた運動中の剛体の姿勢推定

抄録

We proposed a method for estimating the rotation of a moving rigid body in the field of view using high-speed visual information from an event-based vision. We employed a method based on linear regression from the viewpoint of computation cost as a method of estimating the pose of an object using visual information. While changing the pose of the object in the yaw direction, information on the object's appearance was obtained using an event-based vision. The image was generated by accumulating information from an event-based vision for a certain time. By pre-processing these images, binarized images and ternary images were respectively generated. With the above method, a data set of the pose of the object and the appearance of the object was prepared. We verified whether the pose of the object can be estimated from the information of the event-based vision by a linear regression model using dataset.