> For the complete documentation index, see [llms.txt](https://stage-precision.gitbook.io/grid/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://stage-precision.gitbook.io/grid/camera-calibration/using-the-calibration-object/capturing-calibration-data.md). # Capturing Calibration Data After the Calibration Object is configured and the Capture Window is open, you can start collecting calibration data. The goal is to capture useful samples that describe the real camera, lens, tracking, movement system, and calibration references as well as possible. Good calibration data should cover: * the camera image sensor * the visible Calibration Screens or Measurement Points * the physical shooting volume * the camera movement range * the required zoom and focus range * the relevant tracking or axis movement range {% hint style="info" %} The exact capture process depends on the selected calibration workflow. However, the general capture strategy is similar for most tools. {% endhint %} *** ## Start with a Stable Setup Before collecting many captures, make sure the current setup is stable. Do not start capturing large amounts of data until the camera image, detection, and map values look correct. #### Check the Camera Image Open the **Media Input** view in the Capture Window. Check that: * the correct camera image is visible * the image is sharp enough * the Calibration Pattern is readable * the image is not overexposed or too dark * the image is not cropped, scaled, or processed unexpectedly #### Check Marker Detection Open the **April Analyze Engine** view. Make sure the Calibration Pattern is detected reliably. Detected markers should appear consistently when the camera is not moving too fast and the pattern is visible. If detection is weak, check: * focus * exposure * marker size * viewing angle * motion blur * Calibration Image scaling * screen brightness or contrast #### Check Map Values Use the **Map Capture Controller** to confirm that the expected data is updating. Depending on the workflow, this may include: * position * rotation * zoom * focus * rail / position values * pole / lift values * other custom data {% hint style="warning" %} Do not continue capturing if the map values are not updating correctly. Fix the map setup before collecting calibration data. {% endhint %} *** ## Use Auto Calibrate Most workflows use **Auto Calibrate** to collect and process calibration data. When Auto Calibrate is enabled, Grid Studio can capture, analyse, and solve automatically while you move the camera through the required range. #### Enable Auto Calibrate Enable **Auto Calibrate** when you are ready to collect data. Move the camera slowly and deliberately so the system can capture useful samples. Avoid sudden movements, fast pans, heavy motion blur, or views where the Calibration Pattern is barely visible. #### Pause Auto Calibrate Disable **Auto Calibrate** when you want to inspect the current result. After disabling Auto Calibrate, wait until the process states show that capturing, analysing, and solving are idle. This makes sure you are evaluating the latest completed result. {% hint style="info" %} For evaluation, turn **Auto Calibrate** off and wait until **Capture**, **Analyse**, and **Solve** are idle. {% endhint %} *** ## Capture Broad Coverage A good calibration needs diverse captures. Do not collect all data from one camera position or one viewing angle. #### Move Through the Shooting Volume Move the camera through the area where it will be used. Capture data from different: * positions * heights * distances * viewing angles * rotations * movement extremes This helps the solver understand the relationship between the real camera image, the tracking or movement data, and the calibration references. #### Include the Boundaries Capture data near the outer boundaries of the shooting area. This is important because calibration errors often become more visible at the edges of the movement range. Include positions where the camera will realistically be used during production, not only the center of the stage. {% hint style="info" %} Broad coverage across the real shooting volume is usually more valuable than many similar captures from one location. {% endhint %} *** ## Capture Sensor Coverage Sensor coverage describes how much of the camera image sensor is covered by useful calibration data. For lens calibration, this is especially important because distortion can vary across the image. #### Cover the Full Image Area Try to capture Calibration Patterns across the full camera image. This includes: * center of the image * left and right edges * top and bottom areas * image corners If all detected markers are only in the center of the image, the calibration may not have enough information about the outer sensor area. #### Use Different Distances and Angles Move closer and farther away from the Calibration Screens. Use different viewing angles so markers appear in different parts of the image. This helps improve sensor coverage and gives the solver more useful information. *** ## Capture Screen Coverage Screen coverage describes how much of the Calibration Screens has been observed. For screen-based workflows, the solver needs enough observations of the screens used in the calibration. #### Observe All Required Screens Make sure all Calibration Screens used by the workflow are visible in the captured data. If a screen is never seen or only seen from one poor angle, the calibration result may be weaker. Capture each relevant screen from different positions and angles. #### Capture Missing Screen Areas Use the Solve Status feedback to identify missing or weak screen coverage. If a screen or screen area is shown as missing, red, or insufficiently covered, move the camera to a position where that area is visible and capture more data. {% hint style="warning" %} If screen repositioning is enabled, poor screen coverage can also affect the calculated screen placement or shape. {% endhint %} *** ## Capture Zoom Data For zoom lenses, the calibration should cover the zoom range that will be used. Prime lenses do not need zoom range coverage. #### Start with Key Zoom Positions Capture data at important zoom positions. A common approach is to start with the maximum zoom position, then evaluate the zoom range and add data where alignment error becomes visible. Useful zoom positions may include: * maximum zoom * minimum zoom * positions where the alignment error is highest * positions that are frequently used in production #### Move While Capturing Zoom Positions At each important zoom position, move the camera through the space. Do not capture only one view at a zoom position. Try to collect useful data from different distances, angles, and parts of the shooting volume. #### Use Zoom Focus Steps Use the **Zoom Focus Steps** view to check which zoom positions have already been captured. If parts of the zoom range are missing or weak, capture additional data at those positions. {% hint style="info" %} For zoom lenses, capture quality is usually improved by adding data at lens positions where the real and virtual views show the largest difference. {% endhint %} *** ## Capture Focus Data If the lens calibration needs to cover different focus positions, capture additional data across the required focus range. #### Start from Infinity Focus For full lens calibration, start with infinity focus. This usually gives a stable initial lens solve and helps keep distant Calibration Screens detectable. #### Add Stable Focus Positions After the initial calibration, rack focus back from infinity and observe marker detection. Capture additional focus positions where the Calibration Pattern is still detected reliably. Avoid focus positions where markers become too blurry or detection becomes unstable. #### Do Not Guess Unstable Focus Positions If marker detection is unreliable at a focus position, do not collect large amounts of data there. Return to the last stable focus position and capture from there. {% hint style="warning" %} Blurred markers can reduce detection quality and may produce weak calibration data. Only capture focus positions where the Calibration Pattern can still be detected reliably. {% endhint %} *** ## Capture Movement Data Tracked, rotation-only, and axis-based workflows require movement data to match the captured camera image. The capture should cover the movement range that will be used during production. #### Fully Tracked Cameras For fully tracked cameras, capture data from different positions and rotations. Try to cover: * left / right movement * forward / backward movement * height variation * pan / tilt / roll changes * near and far distances to the Calibration Screens #### Rotation Only / PTZ Cameras For rotation-only or PTZ workflows, capture different pan and tilt angles. The camera may not move through 3D space, but the solve still needs enough viewing variation. Capture data across the actual pan and tilt range that will be used. #### Axis 2D Systems For Axis 2D workflows, move through the rail / pole / lift range. Capture different combinations of both Axis 2D input values. The goal is to let Grid Studio learn how the real system moves in 3D space. Capture: * low and high pole / lift positions * different rail / position values * corner or boundary positions of the movement range * positions where the rig bends, twists, or behaves differently *** ## Manual Marker Captures Manual Marker workflows use Measurement Points instead of automatically detected Calibration Patterns. The capture strategy is slightly different because points are referenced manually after the capture. #### Capture Clearly Visible Points Capture images where Measurement Points can be clearly identified. A good capture should contain several accurately measured and clearly visible Measurement Points. Avoid points that are blurry, hidden, ambiguous, or difficult to identify. #### Points Can Differ per Capture Not every Measurement Point needs to be visible in every capture. For moving camera workflows, each capture can contain a different set of visible points. This is expected. The important part is that each referenced point is placed accurately in the image. {% hint style="info" %} For Manual Marker workflows, fewer accurately referenced points are usually better than many uncertain points. {% endhint %} *** ## Evaluate While Capturing Use the Capture Window feedback to decide what data is still missing. Do not rely only on the number of captures. A high number of captures does not automatically mean the calibration is good. #### Watch Reprojection Error Reprojection error can help identify whether the solve is improving. Lower error is usually better, but it should not be the only quality indicator. Also check: * sensor coverage * screen coverage * zoom / focus coverage * movement coverage * visual alignment in the Solve Engine Camera View #### Watch Coverage Coverage feedback helps you decide where to capture more data. Capture additional data where: * sensor coverage is low * screen coverage is incomplete * zoom / focus areas are missing * movement range is not covered * red areas indicate insufficient observations #### Check Visual Alignment Use the **Solve Engine Camera View** to compare the real camera image with the virtual calibration geometry. The result does not need to be perfect during early capture, but it should improve as more useful data is collected. If the solve becomes worse, pause and check for bad input data, wrong maps, poor detection, or incorrect references. *** ## Avoid Bad Captures Bad captures can reduce the quality of the calibration. Avoid collecting data when the image, tracking, or references are not reliable. #### Common Bad Capture Conditions Avoid captures with: * strong motion blur * out-of-focus Calibration Patterns * overexposed or underexposed images * very small or unreadable markers * strong reflections on screens * incorrect Calibration Images * cropped or scaled image input * wrong tracking map * frozen tracking values * incorrect zoom or focus values * ambiguous Manual Marker points #### Stop and Fix the Cause If many captures are bad, do not continue collecting more data. Stop Auto Calibrate, fix the cause, and then continue. {% hint style="warning" %} More data does not fix bad data. If the input image, map values, or references are wrong, the calibration result will also be wrong. {% endhint %} *** ## When to Stop Capturing Stop capturing when the calibration result is stable enough for the required workflow. The exact point depends on the setup, but the result should be evaluated across the full intended range. #### Good Stop Conditions You can usually stop capturing when: * marker detection is reliable * reprojection error is stable * sensor coverage is sufficient * screen coverage is sufficient * required zoom positions are covered * required focus positions are covered * movement boundaries are covered * visual alignment looks correct in the Solve Engine Camera View #### Continue Capturing When Needed Continue capturing if: * coverage is missing * red areas remain in the solve feedback * alignment error is high at specific zoom or focus positions * the result drifts near movement boundaries * the result only works from one camera position * the solve is unstable *** ## Before Continuing Checklist Before finishing the capture process, check the following: * Auto Calibrate has collected useful captures. * Capture, Analyse, and Solve are idle before evaluating the result. * Calibration Pattern detection is stable. * Map values are updating correctly. * Sensor coverage is sufficient. * Calibration Screen coverage is sufficient. * Required zoom positions have been captured. * Required focus positions have been captured. * Required movement range has been captured. * Manual Marker references are clear and accurate if used. * The visual alignment looks correct in the Solve Engine Camera View. * No obvious bad input data is being used.