super accurate
camera calibration
turnkey
Calibration Station
Main Features
- Calibrates any type of optics
- Captures millions of reference points per pose
- ML-based filtering of corrupted data for the best calibration results
- Exports to industry-standard formats such as OpenCV, or to a pixelwise view-ray lookup table
- Advanced quality assessment tools
- Suitable for R&D, DIY projects, and small-batch productionMetrological calibration completed at identical positions, in the identical environment
The station is built on the active target method: a flat screen displays a sequence of coded patterns, the camera captures them in sync, and the system decodes the images. This approach reveals minute details of the camera's imaging geometry. You can use the raw decoded data for your own calibration pipelines or apply our ML-based geometric and colorimetric calibration.
What the Station Includes
- Flat pre-calibrated screen
- Mac Mini or similar PC
- Preinstalled Radiant Metrics Calibration Studio
- A camera for testing and maintenance
Rich Data Collection
- A 3D reference point is determined for every camera pixel
- Each point comes with an unbiased uncertainty estimate
- Per-pixel photometric sensitivity characterization
- For a 12 MP camera, up to 12 million reference points per pose
We use cosine patterns to encode each monitor pixel, providing a dense set of reference points for the camera pixels.
ML-Based Geometric Calibration
- Supports any camera model: OpenCV, ROS, Halcon, custom free-form (pixelwise view-ray LUT), etc.
- Supports any optical design: perspective, fisheye, catadioptric, telecentric, and more
- Quantified consistency, reliability, and repeatability: automated outlier rejection, train/test split, k-fold cross-validation
- Smart model selection: the system chooses the optimal camera model based on task-specific tolerances and detects over/underfitting
A comparison between traditional checkerboard calibration and our cosine-pattern approach for high-quality industrial cameras with central optics.
A comparison between OpenCV’s calibration model and our free-form pixelwise view-ray lookup table shows significant performance improvements for wide-angle cameras. The example below uses a Raspberry Pi wide-angle camera.
Extended Quality Control
- Data quality: visual maps of decoding uncertainties
- Model-to-data consistency:
- RMS reprojection errors (RPE)
- RPE distribution maps
- RMS forward projection errors (FPE)
- FPE distribution maps
- Model reliability: RPE/FPE measured on test data
- Model repeatability: maps of expected FPE — uncertainty of each pixel's view ray at a given distance
Example re-projection error in each camera pixel.
Example forward re-projection error for each camera pixel depending on the distance.
Example expected forward re-projection error for each camera pixel depending on the distance.
for whom
Robot Developers
Full FOV utilization:
Uniformly accurate SLAM and triangulation in the entire image of wide angle cameras.
Precise 3D information for AI:
well-aligned real-world and model imaging geometries facilitate faster training and more accurate inference.
Stereo:
Support for excellent stereo caliration.
Re-calibration:
Calibration can be performed in a lab on in the field, with inexpensive hardware.
Smartphone Manufacturers
Better image undistortion:
Accurate undistortion in frame corners enables the end user to capture more details in a larger field of view.
Reduced pixel registration errors:
Accurate imaging geometry enables seamless zoom across multiple cameras.
Non-uniform color calibration and compensation:
Matching color sensitivity profiles across multiple cameras.
Industrial OEM Suppliers
Full FOV utilization:
Uniformly accurate triangulation in the entire image of wide angle cameras.
Replacement of expensive industrial cameras:
Commodity cameras with an accurate calibration can deliver comparable quality for 3D applications.
Universal and flexible approach:
Our solution replaces static targets with consumer-grade monitors, supports various optics, and adapts calibration costs to each task. It delivers both photometric and geometric calibration with detailed quality metrics.
Robot Developers
Full FOV utilization:
Uniformly accurate SLAM and triangulation in the entire image of wide angle cameras.
Precise 3D information for AI:
well-aligned real-world and model imaging geometries facilitate faster training and more accurate inference.
Stereo:
Support for excellent stereo caliration.
Re-calibration:
Calibration can be performed in a lab on in the field, with inexpensive hardware.
Smartphone Manufacturers
Better image undistortion:
Accurate undistortion in frame corners enables the end user to capture more details in a larger field of view.
Reduced pixel registration errors:
Accurate imaging geometry enables seamless zoom across multiple cameras.
Non-uniform color calibration and compensation:
Matching color sensitivity profiles across multiple cameras.
Industrial OEM Suppliers
Full FOV utilization:
Uniformly accurate triangulation in the entire image of wide angle cameras.
Replacement of expensive industrial cameras:
Commodity cameras with an accurate calibration can deliver comparable quality for 3D applications.
Universal and flexible approach:
Our solution replaces static targets with consumer-grade monitors, supports various optics, and adapts calibration costs to each task. It delivers both photometric and geometric calibration with detailed quality metrics.
why choose us
High‑Density, High‑Quality Data
Up to 33 mio. calibration points (for an 8k camera sensor) capture even subtle optical nuances with exceptional detail
Quality Improvements
Significant RMS errors reduction compared to the classical checkerboard approach.
Efficiency for Complex Tasks
Free-form LUT-based models streamline ray tracing, triangulation, and other expensive operations in 3D.
ML-Driven Robustness
Machine-learning inspired calibration ensures stable, consistent outcomes.
Arbitrary Types of Optics
Calibrates regular perspective, wide-angle, fish-eye, and aspheric lenses.
Maximal Sensor Utilization via Uniform Calibration Accuracy
Delivers more uniform localization uncertainty across the entire sensor, including edges and corners. Enables up to 100% use of the sensor area for demanding tasks.
Photo & Colorimetric Calibration
Extended method of active targets re-uses same dataset for geometric and photometric calibration. For each pixel we calibrate the sensitivity curve, including the lens effects and non-uniformities of the sensor matrix.
Stereo Camera Calibration
Simplifies and improves the calibration of multi-camera setups for depth and 3D applications.
use cases
Pixel Registration for Multicamera Smartphones
click here
- When the user zooms in on an image, the system switches between cameras. Discrepancies in geometry and color response between cameras lead to “jumps” and sudden color changes.
- Proper calibration of all cameras (both geometric and colorimetric) enables efficient correction and compensation schemes.
- Our calibration significantly reduces errors compared to the state of the art and requires no expensive equipment.
Accuracy of Robotic Gripping
click here
- When a robot arm grips objects based on camera input, positioning errors increase near the boundaries of the working area due to poor calibration quality. As a result, the robot may miss or damage the objects.
- Our free-form model calibration ensures uniform accuracy across the entire frame.
OEM Solutions for Computer Vision
click here
- Optical designers often use industrial cameras and high-end optics for demanding computer vision tasks. Costs can become significant, especially for multi-camera setups.
- Mass-produced cameras with inexpensive (e.g., plastic) lenses may rival such systems in image quality but typically exhibit greater distortion.
- By calibrating a free-form model that accurately captures these distortions, we can achieve performance equivalent to high-end systems in the final application.
FAQ
[open]
IS YOUR METHOD SCIENTIFICALLY PROVEN?
Yes, it is. As the starting reference, please see the following publication:
A. Pak, S. Reichel , J. Burke. Machine - learning - inspired workflow for camera calibration, Sensors 22(18), 6804 (2022).
[open]
Why do I need better camera calibration?
If you use a camera in any computer vision application, the calibration quality determines how accurately your computer perceives the real world. Moreover, if your camera settings or environment conditions change, it makes sense to regularly re-calibrate the cameras. Typical calibration workflows are cumbersome and prone to various errors and biases; we aim to remove unnecessary hurdles and provide access to top-quality calibration to anyone with a browser.
[open]
Can I calibrate a smartphone camera?
Probably. Many smartphones may be configured to behave as regular cameras, and will thus work just fine with our method of active targets. However, some premium smartphones may under the hood try to “improve” every picture by artificially painting or “hallucinating” some image details. This step may involve complex and undocumented algorithms that will interfere with our decoding logic, leading to poor-quality datasets.
[open]
Can I calibrate a DSLR camera?
Yes, but remember that changing any lens settings (F-number, zoom, or focus) may invalidate the calibration results.
[open]
Can I use a Full-HD display for calibration?
Yes, the resolution of 1920x1080 pixels is sufficient for high-quality calibration. A more important property, however, is the monitor’s size: in all calibration images, the displayed pattern must occupy the entirety or a significant portion of the camera’s field of view, while the resulting images should be free from excessive “moire” effects. For some cameras, this is easier achieved with high-density (4K or 8K) monitors.
[open]
Can I use a curved monitor?
In the current version of our software we only support flat rectangular screens. However, solutions using curved gaming screens are known and could be implemented in our framework if we see sufficient demand for this feature.
still have questions?
RADIANT METRICSRADIANT METRICS
[01]
email
[02]
phone