(Original text was formatted as an email, so bear with me here)
Hello!
I’m an animation student from Missouri State University. I’ve been working with a team of my peers for the last few months on our senior thesis project - an interactive virtual reality experience titled The Bell Garden. This project aims to blend digital footage with 3D data captured using the Matterport camera to create a unique and immersive virtual adventure! We’ve teamed up with a company called Oculum Works and captured complete Matterport scans of a local bed and breakfast - this being the basis for our story’s location. I’ve taken on the task of processing this point cloud data and converting it into a workable format for use in either the Unreal or Unity game engines.
I’ve tried several methods (and used a variety of software - 3DS Max, CloudCompare, MeshLab) to reconstruct the point cloud data into mesh structures, but haven’t had fantastic results. The meshes are often jagged/undetailed - despite the high quality and density of the original point cloud data. Textures have also been an issue, as they rarely carry over once the mesh has been constructed.
I wanted to see if Autodesk has ever been approached with these issues, and whether or not anyone has any advice about additional software/techniques I could use to produce more accurate and detailed [textured] meshes? We’ve thought about simply importing the original point cloud data into our game engine - but tests I’ve run using this method don’t provide the clearest results either. The imported point cloud itself usually looks decent, but I can’t figure out how to increase the point size in order to fill out more of the unused negative space (which would, theoretically, create a more fleshed out image of the space without the use of meshes). In this case, I was wondering there's any advice on how to increase the point size of the pointcloud while retaining its original textures?
Any thoughts or suggestions would be -greatly- appreciated. I’m not sure many people have utilized Matterport data in this way (hence the lack of online resources concerning these specific conversion methods) - but we hope to document our workflow for future developers who are interested in the process!
