Three-dimensional reconstruction of hard surfaces
Reflective metal, glass, clear parts, dark plastics — the objects conventional 3D vision fails on
The problem
Structured light projects a pattern and reads its deformation; stereo matches texture. Both assume the surface reflects light diffusely.
Specular metal reflects the pattern elsewhere. Glass lets light pass through. Dark plastics return almost no signal. Worse, interreflections between parts and bin walls generate surfaces that do not exist. Depth reconstruction of non-Lambertian surfaces is still treated as an open problem in robotics.
The cost is usually not slightly worse accuracy — it is a stalled line: when no grasp pose can be computed and no alternative target is in the bin, production waits for a human.
How we approach it
LUC-VISION™ derives depth from FMCW coherent measurement. The criterion is the frequency difference between the return and the local oscillator, not how the surface scatters light back. Whether a surface is specular or matte, textured or not, no longer decides whether it can be imaged.
Where it applies
Mixed-material picking, inspection of reflective solder joints and metal housings, service robots facing glass doors and stainless steel.
Related cases
Dent detection on glossy automotive paint
786,432 点
Ornament capture on reflective bronze
960,420 点
Geometry and color captured together on a painted figurine
1,787,235 点
Full-body human scan
100,000 点
RGBD capture of a plush surface
674,633 点
Ready to Transform Your Application?
Partner with us to integrate cutting-edge photonic sensors into your systems