We actually have done pose estimation by calculating the field relative heading of the turret and in the other hand calculating the perpendicular distance to the target. by calculating these two you can use them to find your X and Y offsets relative to the target and since you already know the position of the target you can easily calculate where you are at on the field.
This year we decided to have our swerve turn roughly to the correct angle based on odometry without vision and then fix the error with the relative yaw error, which worked fine for us. If you would like a more general and robust implementation, we did try using Wpilib’s SwervePoseEstimator class that integrates the vision readings with the current pose based on the encoders (the code commented out here is relevant for you). A version for tank drive can be found here.
Good luck! and if you need some more info I can also send our 2020 code (turret with tank drive).