A laser stripe's image position is pixel information, not a physical workpiece dimension. A fixed camera and line-laser triangulation setup needs a geometric relationship between pixels, camera rays and the laser plane. For a finished sensor that already outputs measured cross-section contours in engineering units, check the manufacturer's calibration and interface definitions before applying a separate pixel model.
From an image to geometric points
Extract stripe positions within the valid image region and record saturation, occlusion, missing segments and multiple reflections. Subpixel extraction describes a position-estimation method; it does not prove dimensional accuracy. The manufacturer's laser-triangulation documentation distinguishes extraction results from calibrated metric-space data.
A calculation path for fixed geometry
| Step | Inputs and processing | Required checks |
|---|---|---|
| Camera calibration | Intrinsic and distortion parameters | Resolution, lens and focus state |
| Laser-plane calibration | Known geometry and stripe observations | Plane pose relative to the camera and coordinate units |
| Ray-plane intersection | Corrected image point and plane model | Invalid points, near-parallel geometry and working range |
| Coordinate transformation | Camera coordinates to workpiece references | Mounting relationship and reference definition |
| Dimension calculation | Valid cross-section contours and geometric fitting | Measurement region, fitting rules and invalid states |
With the camera optical center as the coordinate origin, write the pixel ray as X = t r and the laser plane as n · X + d = 0. Their intersection is X = (-d / (n · r)) r. The ray and plane must not be parallel, and the ray and model parameters must use the same coordinate system. See the OpenCV camera-geometry documentation for generating rays after distortion correction. Select the actual calibration implementation for the physical setup.
Check dimensions using independent geometry
Do not use calibration-fitting data as the entire acceptance dataset. Retain independent reference objects or positions and check dimensional differences and repeated measurements across the agreed working range. Record image reprojection errors separately from physical dimensional errors; the former does not directly establish complete-machine accuracy.
Preserve invalid conditions with versions
Assess calibration validity after changes to the lens, focal length, mounting, image scaling or active region. Record calibration files, algorithms, units and device versions. Report an explicit invalid state when the stripe or geometric intersection is invalid, rather than presenting a previous frame's value as a valid current measurement. This article explains a calculation method and reports no completed device calibration or accuracy test.
Related services: 3D vision and point-cloud processing, project consultation.