Calibration and validation procedures are critical to verifying inspection and ensuring the long-term performance of a machine vision system. Furthermore, manufacturers often have customers who require proof of annual calibration as part of their product acceptance contracts.
Calibration
Consistent and accurate color measurement across images collected over time is essential for reliable object segmentation and defect detection using color-based analysis. Accordingly, regular color calibration and correction is necessary to obtain and implement model tuning parameters. This section will discuss:
- lighting profile correction
- color correction
- spatial calibration
- multi-camera calibration
Lighting Profile Correction
Even in a controlled lighting environment, unbalanced lighting and shadows occur and can contribute to false defect detections and, ultimately, to unwarranted product rejection.
Implementing a lighting profile correction prior to defect detection eliminates intensity variations that commonly occur through system drifts, while ensuring that the integrity of key image features remains intact. Lighting profile correction is one of many image analysis techniques that can be used for robust illumination-invariant defect detection.
ProSensus has expertise with implementing lighting correction techniques for a variety of machine vision system configurations including line scan and area scan camera applications.
Our lighting profile corrections are tailored to each specific application in order to deliver reliable metric quantification and consistent inspection capabilities.
Color Correction
In order for machine vision solutions to deliver accurate and repeatable color-based results and analyses, a color calibration process is typically required.
A color calibration makes use of a standardized Color Checker for consistent color reproduction across imaging systems. Color correction factors are generated to adjust each color chip of known reference values under any illumination, and these factors are then used to train and tune models for a “reference factory state” to ensure consistency of model results.
Using the Color Checker, our machine vision experts can measure and analyze any differences in color reproduction to identify any potential system drifts and to recalibrate the system to ensure reproducible results.
Spatial Calibration
When using a combination of line- and area-scan cameras, spatial calibration is required in order to translate 2D pixel coordinates to real world measurements, and this is accomplished by taking into account line rates and lens distortions.
When angled area scan cameras introduce perspective distortion, perspective correction is completed via coordinate transformation. Automatic spatial calibrations implement perspective correction methodology utilizing chessboard pattern systems.
The chessboard is used to determine the appropriate transformations necessary to output a distortion-free image, rendering the system versatile to installation angles of mounted cameras for consistently accurate defect detection and object size estimation.
Multi-Camera Calibration
Often the lighting conditions, camera settings, and viewing distances may be different for each camera in a multi-camera machine vision system. Yet, the collected images are expected to provide holistic and consistent defect detection. The variability in camera set-ups necessitates active calibration procedures to ensure color and spatial uniformity across all images collected.
Application: Multi-Camera White Tile Calibration
For ProSensus’ 6-camera BaleGuard system, color consistency is important as a single defect detection algorithm is applied to all 6 images belonging to the 6 geometric sides a single rubber bale. A multi-camera calibration equalizes the light response across images with respect to a standard white tile as a calibration reference.
This online procedure ensures reproducibility with accurate bale color temperature, and restores overall image brightness and weighting of individual color channels.
Validation
ProSensus has created system prerequisites for on-going re-calibration of our color machine vision systems to identify significant deviation from the ideal state that should be restored for accurate image analysis and relevance of tuning parameters.
In this image, the Color Checker is used to compare presently captured states to original factory settings (on which model was tuned). Reference factory settings are displayed as the outer edge of each color tile, while presently captured states are the inner values. Each color tile includes error values (above a threshold) for each color tile, indicating deviation from factory settings.
As a result of the validation procedure, a color correction can be applied to the collected images. This image compares color tile of the factory settings to the newly corrected image.
Calibration Certificates
An annual calibration certificate includes documentation of all calibration activities performed, including any detected system drifts and applied corrections. Ensure your long-term system performance and meet your customer-requested proof of annual calibration by contacting ProSensus about our machine vision calibration certifications.
