EDIT - SOLUTION: the paper itself says which vector should be selected as reference vector, the one which gives the maximum positve distance
Due to my reputaion I can not attach all the necessary photo here so I just attached one single dropbox link
I am trying to implement an algorithm below:
I almost completed it only issue is that, the method described is working for single point correctly but when it comes to defining an error band for detection of object in noisy data we need to define a error band on the basis of neighbouring points and while I am doing that I am getting error because everytime the reference vector with which profile is calculated, is changing therefore the profile is always shifted.
I need a way to find a constant vector with which I can calculate the angle so that the profile shift is eliminated.
I am using dot product method to calculate angle between vectors.