Add support for tracking via the main camera. The main motivation is more accurate centering of the object in the field of view, both to prevent it from leaving the camera frame and to potentially allow for capturing a smaller region of interest at a higher frame rate.
There are a few challenges to make this work, but I think they all might have solutions:
Capture from the main camera has historically not been a part of the track package, and it probably ought to remain that way. However for capture software we control it would be possible to use something like ZeroMQ pubsub to send frames (possibly downsampled) from the capture process to the track process efficiently and quickly.
Computer vision algorithms that work for the guidescope probably won't work as well for an extended object with resolved details.
If a guidescope is still used for initial acquisition there will need to be some kind of state machine or sensor fusion to make use of both cameras.
Add support for tracking via the main camera. The main motivation is more accurate centering of the object in the field of view, both to prevent it from leaving the camera frame and to potentially allow for capturing a smaller region of interest at a higher frame rate.
There are a few challenges to make this work, but I think they all might have solutions: