Hi guys, I am new to depth cameras and lidar-based 3d scanning, but have some experience from photogrammetry and Meshroom.
I recently came into an Intel L515 that I’ve been getting decent results from. I read through the Sensor Choice FAQ, but still had some questions. My use-scenario has been to build a mechanically accurate meshes that I can manually build into a boundary-represented model in a CAD program. It would be nice if the sensor/software combination could discern things like clearance holes for threaded fasteners in the range of 0.5" and smaller.
For the L515, I’ve had to insert rolled pieces of paper in holes to obtain good location data on the central-axis and size of the holes I’ve been trying to measure.
I was leaning towards the D435i because it seemed like the combination of short scanning range and the inertial feedback would lead to acceptable texture results in combination with decent accuracy working across objects that would be in the size range of bicycle to passenger car.
Lastly, what are some strategies I can employ for improving the overall confidence of the depth measurements? Is setting up fiducial targets/standards a waste of time compared to dusting glossy surfaces with baby powder or baking soda?
Alternatively, if I wanted to rapidly convert real world objects into VFX or videogame assets, would there be a different choice? It seems like the Kinect Azure is the hands down best choice for high-quality texture capture, but given the scarcity of Kinect Azure SDK kits it sounds like my next best choice is photogrammetry if texture quality is my top priority.