Uhm... A bounding box is the rectangle with smallest area containing what you want. (points, cicarcle, etc.)
It can be generalized to be a box-like container which has the smallest measurement containing some arbitrary number of elements.
You ask how to use them. Do you mean that you have already created a bunch of rectangles and now want to work out collision-detection with them? Is this correct?
The naive way is simply to check if any box overlap another box. You can use an initial check and then only check boxes when they move.
You can have a grid of some arbitrary size, keep check of which boxes are fully or partly in which parts of the grid. Then you only have to check the relevant parts of the grid for collisions.
I hope this answered at least partly what you asked.
*hugs*