I'm sorry if this is not the right SE site for this question...
background: I have a kalman filter in which some of the measurements are points in some 2d plane. in certain occasions I know (or rather, have some prior knowledge) 1 of my points is lying on some line.
while normally I would like to model each point x and y coordinates as independent normal random variables, thus my measurements covariance matrix is diagonal in some occasions I would like to express something among the lines of "the point consisting of x and y should be located on some known line with some additional noise". I assume I should be able to express it as covariance between x and y but I would like some help regarding how
thanks