I think the core of your issue is you chose an algorithm that's not well suited to solving the problem. The algorithm is fast, but it has an achilles heel with respect to errors where the line of intersection passes exactly through a vertex of the mesh you're cutting. The algorithm is also dependent on screen resolution. If you have a low resolution screen (viewport) in the area being tested, then expect inaccuracies. Given the oblique angle of the camera relative to the side of the mesh exhibiting the issue, and the fact your mesh is quite dense relative to those pixels, it looks to me that is the problem. While it's always preferred to do things in the speediest fashion possible, your immediate task doesn't require that kind of speed.
I had to do a similar task last week. I prototyped my solution in JScript for convenience, and using a classic ray-plane intersection test, converting all mesh vertex positions to global space, testing every triangle on the mesh, and so on. Despite all that extra overhead, I still got performance good enough for your problem. Convert that to C++ and do some optimization by reducing which triangles to test, etc.. you should have no problem with a classic ray-plane intersection algorithm. Alternatively, you could do line intersection tests on the polygon edges instead. That may be a bit faster, but a little more hassle as once you find the edges, you have to do a little work to find which polygons they belong to. To determine which polygons are facing the camera....if you know how the supercover algorithm works, you shouldn't be asking that question. ;-) Matt Date: Sun, 30 Apr 2017 10:24:11 +0300 From: Andrew Prostrelov <prostre...@gmail.com> Subject: Re: Another rip question To: "Official Softimage Users Mailing List. https://groups.google.com/forum/#!forum/xsi_list" <softimage@listproc.autodesk.com> Oh. I know this thread ;). It gave me idea of cutting by parametric plane and not by vector projections. But unfortunately there are no answer for my question in this thread and i don't have any ideas left. By the way there is another interesting question: how should i differentiate visible side of mesh components and non visible for current Camera position. http://stackoverflow.com/questions/9709970/algorithm-or-software-for-slicing-a-mesh So ... CGAL maybe. I don't know does some one have an experience of using CGAL lib ? ------ Softimage Mailing List. To unsubscribe, send a mail to softimage-requ...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.