I am trying to land very precisely on a stationary visual target using velocity commands. I would like to improve the accuracy of the EKF’s velocity estimation by calculating the velocity of the target in the camera frame and sending it to the FCU via a mavlink message. It seems like VISION_SPEED_ESTIMATE is the best message for this. According to the mavlink documentation, the message expects the global x, y, and z speed as parameters. Does this mean lat/long? Should the speed be expressed in m/s or deg/s? If no messages are being sent (target not in sight), will the EKF ignore the vision estimate?
Interested to know more about VISION_SPEED_ESTIMATE!