Single Camera + Cross-Line Laser: Dense 3D Point Cloud from One Image
Hello everyone,
I would like to share an experimental 3D reconstruction project and ask for technical feedback from people working in computer vision and 3D reconstruction.
The hardware setup is very simple:
-
One monocular camera
-
One cross-line laser projector
After calibration, the system captures a single image containing the object and the projected laser cross.
In conventional laser triangulation, 3D coordinates are typically obtained only for image pixels lying on the observed laser profile. My approach attempts to extend this: after calibration, the reconstruction is not limited to the laser-profile pixels. In principle, 3D coordinates can be calculated for all pixels across the image plane, producing a dense point cloud from a single image.
Here is the basic idea:
The four reference directions provided by the cross-line laser are used as geometric constraints. From these constraints, I construct a continuous mapping between image coordinates and corresponding 3D geometric directions. The resulting camera ray and reconstructed laser-related ray are then intersected to obtain the 3D position.
The current prototype has already produced dense point clouds from a single captured image.
Demo video:
What I would especially like to know is:
-
Are there existing reconstruction methods similar to this approach?
-
Is there known work on recovering a continuous 3D direction field from a small number of laser reference lines?
-
From a computer vision perspective, what would be the most important experiments or evaluations needed to verify such a method?
I am not claiming that this replaces established multi-view reconstruction methods. My goal is simply to explore whether a calibrated monocular camera and a cross-line laser can provide enough geometric information for dense 3D reconstruction from a single image.
I would greatly appreciate technical criticism, references to related work, or suggestions for validation experiments.
Thank you.