opticalflow - How to use the extended kalman filter for IMU and Optical Flow sensor fusion? -
i building quadcopter , using pixhawk autopilot system px4flow sensor attached optical flow data. px4flow high speed smart camera (arm processor) integrated gyro , height sensor, , outputs linear velocities internal optical flow algorithm.
now, improve on position , velocity estimates using extended kalman filter fuse imu , optical flow data. have derived state model function , state transition matrix prediction step.
by problem lies within deriving measurement model/function optical flow velocities, used in update phase of extended kalman filter. believe have derive optical flow algorithm how, far have gotten.
*edit: here article describing px4flow unit , how calculates velocities. (forgot add link, it's there)
i'll assume (since don't give details) state vector contains speed , position in world coordinates. optical flow returns xy speed in body coordinates, have rotate according yaw angle speed in world coordinates.
if case, can choose between:
- if state vector not include orientation, pre-transform optical flow speed world coordinates, , use linear observation model (zero matrix ones in position of x, y speed in state vector)
- if state vector includes orientation, can write full observation model transform between world-body coordinates using angles.
that's process. however, better answer should provide more details.
Comments
Post a Comment