What is a good alternative to RANSAC
I have a cloud of points that I am fitting a plane of best fit. This cloud of points is representing the ground in front of my vehicle. When I am sitting still, there is too much randomness and the plane of best fit jumps around too much. I am currently using RANSAC to find my plane of best fit. I need a more consistent or deterministic method that is still efficient. I understand that I will not be able to maintain the same efficiency but any ideas would be great. Essentially when I sit still I need the same (or almost the same) plane to be generated.
you can try out a GraphCut RANSAC proposed by Daniel Barath and Jiri Matas at CVPR 18' Paper and GitHub. The approach is implemented using OpenCV 3.31 and the method should be more accurate and robust but I havn't test it.