Laserline with Industrial Data

[COLLAB] This chapter requires user collaboration for industrial application context, data collection protocols, and accuracy requirements.

This chapter discusses practical considerations for laser triangulation calibration in industrial settings.

Industrial Context

Laser triangulation is widely used in industrial inspection:

• Weld seam inspection: Verifying weld bead geometry (width, height, penetration)
• 3D surface profiling: Measuring surface topology for quality control
• Gap and flush measurement: Checking assembly tolerances in automotive manufacturing
• Tire and rubber inspection: Measuring tread depth and sidewall profiles

In each case, a laser line is projected onto the workpiece and a camera observes the deformed line. The calibrated laser plane, combined with camera parameters, allows converting pixel positions to metric 3D coordinates.

Accuracy Requirements

Data Collection for Laser Calibration

Calibration Board Setup

The calibration board serves dual purpose:

1. Provides chessboard corners for camera intrinsics calibration
2. Provides a known plane to constrain the laser line observations

Capturing Views

For each view:

1. Position the calibration board at a known angle to the laser plane
2. Capture an image with both the chessboard pattern and the laser line visible
3. Record chessboard corners and laser line pixel positions separately

View Diversity

• Board angles: Vary the board tilt so the laser-board intersection line covers different positions in the image
• Board distances: Vary the distance to cover the working range
• Minimum views: 5-10 views with diverse orientations

The laserline_device_session Example

cargo run -p vision-calibration --example laserline_device_session

This example uses synthetic data to demonstrate the workflow. For real data, the laser pixel extraction must be performed by an external algorithm (typically a peak detector applied to each image column).

Interpreting Results

Key outputs:

• Laser plane normal: The direction the laser projects in. Should align with the physical mounting.
• Laser plane distance: Distance from camera center to the laser plane. Should match the physical geometry.
• Mean laser error: The average residual of laser observations. In pixel units for LineDistNormalized, in meters for PointToPlane.

DS8 Dataset

The data/DS8/ directory contains a real industrial laser triangulation dataset:

• ExperimentDetails.txt: Experiment metadata
• calibration_object.txt: Calibration pattern specification
• images/: Captured images
• robot_cali.txt: Robot calibration data (if applicable)

Troubleshooting

• Poor laser plane estimate: Usually caused by insufficient view diversity or laser pixels too close to collinear (all from similar board angles)
• High laser residuals: May indicate laser pixel detection errors, or that the laser is not well-described by a single plane (diverging laser, curved projection)
• Scheimpflug convergence: If sensor tilt parameters diverge, try initializing with fix_sensor: true for the first few iterations, then release