[Hitchhiker 42]

First, I want to stress, this will not be able to be done in one night. Carefully explain to your team that such a task is a large one that requires plenty of writing, testing, re-writing, etc.

Now to the actual meat:
There are 3 main parts to vision processing/tracking.

1. Finding the vision target and extracting data from it. This can be done using several different vision programs, such as robo realm or GRIP. Our team uses NI Vision Assistant.
2. Processing data from this image and turning it into actual gyro headings/distances. This requires a bit of trignometry, and I suggest taking to a whiteboard or piece of paper to figure this out. Once you draw it out, it becomes a relatively simple math problem.
3. Use these new-found "real-world" values (as oppose to camera values) to adjust the robot to face the right way and be the right distance away. This will require a PID loop. One thing to stress: don't use the camera as the feedback sensor; use the gyro/encoders for this.

Notes:

Step 1 will require a lot of testing to get the program to pick out the target just right. You then will probably need to find at least the COG of the target, but probably also the bounding box. Again, it is easier to use a separate program to develop this code.

Step 2 is pretty self-explanatory.

Step 3 will be quick to set up, especially if you're using CAN Talons with their built-in PID.