Implement bottom-facing-camera velocity detection

[connor-wool]

Using the bottom-facing-camera and machine learning video software (PIDIC?) we would like to measure velocity, or at least if velocity is non-zero or zero.

This will require reading in the image feeds off of the robosub system, learning how to hand these image feeds into our vision system, and how to get the vision system to output a final value for velocity.

This will help improve Kalman Filter accuracy, since more sources of data are (usually) always helpful.