Using FAST corner detector to extract rectangles
I try to make some modifications to "ArUco" augmented reality library (which is written by using OpenV) to make it faster in real time. They are using Canny edge detector to detect fiducial markers, which have rectangular shape. My suggestion to improve the speed of the detection is to use FAST corner detector to detect the fiducial markers (the rectangles) found in the scene in real time.
My questions:
1- Is FAST corner detector is the best choice to improve the speed of the process of marker detection instead of canny edge detector?
2- If it is an appropriate choice, how can I use it to detect rectangles in the scene?
3- According to the process of marker pose estimation (in which the position of the rectangle in the scene is calculated), they used "An algorithm for Least-Squares estimation of nonlinear parameters". What is the perfect alternative to that algorithm to get faster result. Can I use FAST corner detector in this step?
Thank you very much