Determining the importance of real-time visual information for the remote supervision of an autonomous agricultural machine