John Mahoney

Douglas,

Thanks for the nice function.
Could you please clarify the behavior depending on the ROBUST flag?

From the function doc:

% ROBUST (optional) set to 1 or true means to use a slight variation of the
% algorithm that might return duplicates of some intersection points, and
% then remove those duplicates. The default is true, but since the
% algorithm is slightly slower you can set it to false if you know that
% your curves don't intersect at any segment boundaries. Also, the robust
% version properly handles parallel and overlapping segments.

First, based on the first sentence, it would seem that the removal of duplicates is handled by intersections.m. This is not the case.

Second, it would seem that something should fail if the curves *do* intersect at segment boundaries (points specifying the line) and the ROBUST = 0/false is used.
At least for the example below, this is not the case.

Thanks for your help!

% Test intersections function
% Depending on whether ROBUST is true or false, and depending on how points
% defining each line do or don't lay (exactly = machine prec?) on the other
% line, the algorithm returns a different number of intersections.

clear all

% % two lines where no point is exactly on other line
% x1 = [0.1, 0.4, 0.5, 0.8];
% y1 = [0.2, 0.5, 0.7, 0.9];
% x2 = [0.1, 0.4, 0.8];
% y2 = [0.9, 0.3, 0.1];

% % first line has point on the other, but not vice versa
% x1 = [0.1, 0.25, 0.5, 0.8];
% y1 = [0.2, 0.6, 0.7, 0.9];
% x2 = [0.1, 0.4, 0.8];
% y2 = [0.9, 0.3, 0.1];

% % two lines that share a common point
% x1 = [0.1, 0.4, 0.5, 0.8];
% y1 = [0.2, 0.3, 0.7, 0.9];
% x2 = [0.1, 0.4, 0.8];
% y2 = [0.9, 0.3, 0.1];

robust = 0;
[xinter, yinter, iinter, jinter] = intersections(x1, y1, x2, y2, robust);

figure
hold on
scatter(xinter, yinter, 200, 'r','filled')
plot(x1, y1, '.b-','markersize',30)
plot(x2, y2, '.g-','markersize',20)
text(0.5, 0.5, ['found ', num2str(numel(xinter)), ' intersection/s'], 'fontsize',14)