I am using a terrain mesh generated from LiDAR data to build a graphic gallery image for a vehicle simulation and I've tried both obj and fbx exports. When I generate collision geometry for the part I generated from the graphics gallery approximately every other mesh triangle has it's normal reversed.
Some software isn't very strict when applying normal's to vertex data when exported, so I've been running all exports through a mesh viewer to verify mesh normal's are correct before bringing them into Vortex.
I've been generating meshes and exporting obj & fbx files from 3D Reshaper & Sequioa with same results.
I have all of the data transformed to a local csys to eliminate issues potential caused by large offsets.
it would be interesting to confirm that the generated mesh is actually valid by enabling backface culling in your mesh viewer. If you cannot do so, the fbx file, which is usually readable, should provide you with the information. If you get the same result as in vortex, the problem is most probably the winding order of every other triangles instead of the normal. You can also try to change the orientation of the mesh by right clicking the geometry inside vortex and change its orientation. If the hidden triangles switch to the previously unhidden ones and vice-versa, you have definitely have a winding order issue.
I have verified the mesh that I am generating a mesh from in three different viewers. All three viewers have back face culling capability and the mesh that I am using for the graphic gallery image has been verified in all three. The culling is accurate and the mesh normals are visibly correct before I import it into Vortex. I have checked the culling for the mesh in obj and fbx exports.
My current guesses are that our import might not support how the mesh is set in the fbx file (triangle fan?) or maybe it has to do with the coordinate system. The best would be for me to have a look at the mesh. If you cannot, have you looked in the mesh data itself? How is the mesh set up?