Difference between revisions of "Quick Start Guide: Precision Capture"

Line 121: Line 121:
  
 
==Calibration==
 
==Calibration==
Below calibration instructions are specific for Micron volume capture application.
 
 
 
 
===Wands===
 
===Wands===
 
For calibration, we recommend using a Micron Series wand, either the CWM-250 or CWM-125. These wands are made of invar alloy, very rigid and insensitive to temperature, and they are designed for providing precise and constant calibration measurements. At the bottom of the wand head, there is a label tag which shows a factory-calibrated wand length with sub-millimeter accuracy. For best results, input this information into Motive. In the Calibration pane, select Micron Series under the OptiWand dropdown menu, and define the exact length in the wand length section.
 
For calibration, we recommend using a Micron Series wand, either the CWM-250 or CWM-125. These wands are made of invar alloy, very rigid and insensitive to temperature, and they are designed for providing precise and constant calibration measurements. At the bottom of the wand head, there is a label tag which shows a factory-calibrated wand length with sub-millimeter accuracy. For best results, input this information into Motive. In the Calibration pane, select Micron Series under the OptiWand dropdown menu, and define the exact length in the wand length section.
Line 134: Line 131:
  
 
Note: '''Avoid''' touching the marker on the CWM-250 or CWM-125 since any changes can affect the calibration and overall data.
 
Note: '''Avoid''' touching the marker on the CWM-250 or CWM-125 since any changes can affect the calibration and overall data.
 +
  
 
===Calibration Checkpoints===
 
===Calibration Checkpoints===
Line 144: Line 142:
 
*The amount of acceptable error is a complicated subject because the calibration can only use its own samples for validation. Acceptable mean error varies based on your volume size, number of cameras, and desired accuracy. The key metrics to keep an eye on are the Mean 3D Error for the Overall Reprojection and the Wand Error. For our testing we only used calibrations with the Mean 3D Error less than 0.080 mm and the Wand Error less than 0.030 mm. Again, the acceptable numbers are subjective, but generally lower numbers are better.
 
*The amount of acceptable error is a complicated subject because the calibration can only use its own samples for validation. Acceptable mean error varies based on your volume size, number of cameras, and desired accuracy. The key metrics to keep an eye on are the Mean 3D Error for the Overall Reprojection and the Wand Error. For our testing we only used calibrations with the Mean 3D Error less than 0.080 mm and the Wand Error less than 0.030 mm. Again, the acceptable numbers are subjective, but generally lower numbers are better.
 
*Re-emphasizing from previous sections, use a Micron Series wand if possible, and NEVER touch or bump the markers on the micron wand.
 
*Re-emphasizing from previous sections, use a Micron Series wand if possible, and NEVER touch or bump the markers on the micron wand.
 
  
 
==Reconstruction Optimization==
 
==Reconstruction Optimization==

Revision as of 22:04, 13 January 2016

Warning2.png

THIS PAGE IS UNDER CONSTRUCTION THIS PAGE IS UNDER CONSTRUCTION THIS PAGE IS UNDER CONSTRUCTION THIS PAGE IS UNDER CONSTRUCTION THIS PAGE IS UNDER CONSTRUCTION THIS PAGE IS UNDER CONSTRUCTION THIS PAGE IS UNDER CONSTRUCTION THIS PAGE IS UNDER CONSTRUCTION

Quick Start Guide: Precision Capture is under development. It will be available soon. Thank you for your patience!


Back to the Main PageBack to the Quick Start Guides List


This quick start guide details specific instructions for capturing in a small to medium sized volume for extremely accurate tracking using an OptiTrack motion capture system. For more general information not included in this page, please refer to the Quick Start Guide: Getting Started or corresponding workflow pages. Before going into details on how to optimize the system for precision tracking applications, let's briefly go over the residual value, which is the key reconstruction value important for analyzing and monitoring the precision of tracked markers.


Residual Value

The residual value is a distance offset between the converging rays when reconstructing a marker; hence indicating preciseness of the reconstruction. The tolerable residual distance is defined from the Reconstruction pane. If you select one or more markers in the Live mode or from a recorded 2D data, then the mean residual value will be displayed in the status bar located at the bottom of Motive. With a smaller residual value, tracked rays converge more precisely and more accurate 3D reconstruction is achieved. A well-tracked marker will have a sub-millimeter average residual value.


MVQSG Residual1.png MVQSG Residual2.png


Camera Placement

Proper camera placement techniques can greatly improve tracking results. The following guide highlights important setup instructions for the small volume tracking. For more details on the system hardware setup, read through the Hardware Setup pages.


Setup Environment

Avoid populated area when setting up the system and recording a capture. Clear any obstacles or trip hazards around the capture volume. Physical impart on the setup will distort the calibration quality which could be critical especially when tracking at sub-millimeter accuracy. For best results, routinely recalibrate the capture volume.


Mounting Locations

Place the cameras in a random pattern or circumnavigating arrangement around the target volume. For more precise tracking, it is beneficial to place the cameras closer, but avoid placing them right next to each other. Cameras in close vicinity will capture only similar vantages which will not contribute to the reconstruction.

For more information, read through the Camera Placements page.

MVQSG Placement1.png MVQSG Placement2.png


Captured 2D images from each camera are used to precisely locate the 2D positions from the perspective of the camera, but the data on the remaining dimension, in and out of the screen, can only be obtained through reconstructed 3D data from multiple cameras in the setup. Accordingly, for accurate results in all directions, cameras should be mounted so that each camera gets unqiue vantage of the target volume. A spherical arrangement will be more beneficial for tracking all three axis.


MVQSG Placement3.png MVQSG Placement4.png

Mount Securely

For most accurate results, cameras should be securely mounted onto a truss system or an extremely rigid object. For precision tracking applications, any slight deformation or fluctuation to the mount structures may affect the tracking quality. Small size truss system is ideal for the setup. Take extreme caution when using speed rails mounted onto a wall, because building may fluctuate greatly especially on hot days.

For more information on mounting structures, read through the Camera Mount Structures page.

CameraMount Manfrotto.png Mount Clamp2.png

Focus and Aiming

F-stop

Increase the f-stop to a higher number (smaller aperture) to gain a larger depth of field. Increased depth of field will make greater portion of the capture volume in-focus and will make measurements more consistent throughout the volume.

Aim and Focus

Especiallly for precise and close-up captures, cameras aim and focus should be adjusted as perfectly as possible. Aim the cameras towards the center of the capture volume. Optimize the camera focus by zooming into a marker in Motive, and rotating the focus knob on the camera until the smallest marker is captured as clearly as possible with the best image contrast

For more information, please read through the Aiming and Focusing workflow page.


Capture Volume

Infrared Black Background Objects

The motion capture cameras detect reflected infrared light. Thus, having other reflective objects in the volume will alter the results negatively, which could be critical especially for precise tracking applications. If possible, have background objects that are IR black and non-reflective. Capturing in a dark background provides better contrast between bright and dark pixels, which could be less distinguishable in a white background. THe following images show a clear image of a marker with good contrast (left) and a less clear marker (right) whose centroid calculation may have been compromised by an extraneous bright pixels from the background.

MVQSG IRBlack.png 300 px

Environmental Factors

Environmental factors can play a surprisingly large role when measuring at a sub-millimeter accuracy.

Temperature

In a mocap system, camera mount structures and other hardware setup components may be affected by temperature fluctuations. Refer to linear thermal expansion coefficient tables to find out what materials are susceptible to temperature changes. For example, iron has relatively high thermal expansion coefficient, and therefore, you have to be careful when mounting cameras onto iron mounting structures. For best accuracy, take the temperature fluctuation into an account both when selecting the mount structures and before collecting data and routinely recalibrate the capture volume.


Ambient Temperature

An ideal setup is to have temperature controlled capture volume, but if that is not possible there are few considerations to keep in mind.


  • Place the truss or tripods on a rigid surface such as concrete (avoid carpet). Mounting to the walls with something like speed rail can be dangerous as well because many buildings may fluctuate with changing outside temperature. This is especially noticeable on hot days and will significantly affect your results.
  • Aluminum has a particularly high coefficients of linear thermal expansion. So one should be wary when using aluminum based truss.


If these situations are unavoidable, then the solution is to consistently monitor the average residual value for how well your rays converge to individual markers.


Camera Heat

The cameras will also heat up with extended use, and change in internal hardware temperature may also affect the capture data. For this reason, avoid capturing and/or calibrating right after powering the system. Tests have found that the cameras need to be warmed up in Live mode for about an hour until it reaches a stable temperature. For Ethernet camera models, camera temperatures can be monitored from the Camera Preview (2D) pane in Motive (Camera Preview (2D) pane > Eye Icon > Camera Info).


Vibrations

Avoid any vibrations or movements of the setup while taking precision measurements. Especially for measuring at sub-millimeters, even a minimal shift can affect the recordings. In particular, watch out for the following:

  • Never touch the truss or mounting device for the cameras.
  • Make sure your capture area is away from heavy foot traffic and that people aren’t walking around while capture is taking place.
  • Closing doors, even those on the other side of the, may be noticeable during recording.

Re-calibrate the capture volume if your average residual values start to deviate or the capture data starts looking strange.


Micron Capture Motive Settings

The following sections cover some key configuration settings which may need to be optimized for the micron volume tracking.


Cameras Pane Settings

To open the cameras pane, click the camera icon at the top of Motive.

  • Set the Gain setting to 1: Low (Short Range) for all cameras. This will help reduce noise in the image.
  • Set the system frame rate (FPS) to its maximum value. If you wish to use slower frame rate, use the maximum frame rate during calibration and turn it down for the actual recording.
  • Do not bother changing the Threshold (THR) or LED values. Keep them at their default values THR = 200 and LED = 15. The Values EXP and LED are linked so change only the EXP setting for brighter images. If you turn the EXP higher than 250, make sure to wand extra slow to avoid blurred markers.

At this point, you should’ve already adjusted the f-stop for each camera. Now you need to adjust the camera exposure setting for optimized tracking. The next section will go into methods for optimizing tracking quality. For now, set this number as low as possible while maintaining the marker tracking without making them look like gradients (you want them to look like white circles).


Reconstruction Settings

To open the cameras pane click View > Reconstruction Settings the top of Motive. Read through the Reconstruction page for specific details. For the micron volume tracking, important reconstruction settings and the appropriate values are listed below:

  • Residual (mm) < 2.00
    • Allowable residual value should be set smaller for the micron volume tracking, because any offset above 2.00 mm residual value will be considered as inaccurate.
  • Minimum Rays >= 3
    • When more rays converge within allowable residual offset, reconstructed marker will have more accurate position.
  • Minimum threshold pixels >= 4
  • Circularity >= .6


Calibration

Wands

For calibration, we recommend using a Micron Series wand, either the CWM-250 or CWM-125. These wands are made of invar alloy, very rigid and insensitive to temperature, and they are designed for providing precise and constant calibration measurements. At the bottom of the wand head, there is a label tag which shows a factory-calibrated wand length with sub-millimeter accuracy. For best results, input this information into Motive. In the Calibration pane, select Micron Series under the OptiWand dropdown menu, and define the exact length in the wand length section.


MVQSG Wandlabel.png MVQSG WandCalib.png


The CW-500 wand is designed for large to medium capture volumes, but it is not suited for calibrating the micron volume. Not only it has longer reference dimension, but also the wand is made with aluminum, which makes it more vulnerable to thermal expansions. In order to calibrate the volume with best accuracy, the distance between the wand markers must maintain accurate and constant. During the wanding process, Motive assumes the defined wand length, and any distortions would cause the calibrated capture volume to be scaled slightly differently, which can be significant when capturing precise measurements.

Note: Avoid touching the marker on the CWM-250 or CWM-125 since any changes can affect the calibration and overall data.


Calibration Checkpoints

The following are some tips for calibration to get most accurate results.

  • Wand slowly. Waving the wand around quickly at high exposures will blur the markers and distort the centroids reducing the quality of your calibrations.
  • Try not to occlude any of the calibration markers while wanding. Occluding markers will reduce the quality of the calibration.
  • A variety of unique samples is needed in order to achieve a good calibration. For example, if you wand only in horizontal orientation, the calibration may fail. So make sure to wand in a three dimensional volume, put the wand in a variety of orientations, and wand the edges of the volume as well.
  • Wanding the edges of the volume helps improve the lens distortion calculations, but it may also lead to a slightly worse calibration report. This is because the calibration engine creates the report using the collected wanding samples, and the samples collected near the edge of the volume will have more errors which could affect the overall result.
  • The amount of acceptable error is a complicated subject because the calibration can only use its own samples for validation. Acceptable mean error varies based on your volume size, number of cameras, and desired accuracy. The key metrics to keep an eye on are the Mean 3D Error for the Overall Reprojection and the Wand Error. For our testing we only used calibrations with the Mean 3D Error less than 0.080 mm and the Wand Error less than 0.030 mm. Again, the acceptable numbers are subjective, but generally lower numbers are better.
  • Re-emphasizing from previous sections, use a Micron Series wand if possible, and NEVER touch or bump the markers on the micron wand.

Reconstruction Optimization

The following sections covers methods that can be used to check the tracking accuracy and better optimize the reconstructions settings in Motive.


Optimization Method 1

First, go into the perspective pane > select a marker, then go to the Camera Preview Pane > Eye Button > Marker Centroids = True. Make sure your cameras are in object mode, then zoom into the selected marker. The marker will have two crosshairs on it; one white and one yellow. The amount of offset between the crosshairs will give you an idea of how closely the view of that camera (white) aligns with the reconstructed position (yellow). Switching between the grayscale mode and the object mode will make the errors more distinguishable. The below image is an example of a poor calibration. A good calibration should have the yellow and white lines closely aligning over each other.

MVQSG Optimization1.png

Optimization Method 2

The calibration quality can also be analyzed through checking the convergence of the tracked rays into a marker. This is not as precise as the first method, but the tracked rays can be used to check calibration quality of multiple cameras at once. First of all, make sure tracked rays are visible; Perspective View pane > Eye button > Tracked Rays. Then, open the perspective view pane and select a marker. Zoom all the way into the marker (you may need to zoom into the sphere), and you will be able to see the tracking rays (green) converging into the center of the marker. A good calibration should have all the rays converging into approximately one point, as shown in the following image.


MVQSG Optimization2.png

Optimization Method 3

For the micron volume capture, it is not always necessary to set the camera exposure to its lowest value, but instead, the exposure setting should be configured so that the reconstruction is most stable. To test this, zoom into a spot and examine the jitter of a single marker as you change the exposure setting. Use the exposure value that has the least amount of jittering. Methods #1 and #2 can be used for verification.


Optimization Note: Intrusion

When markers are placed very close to each other, their reflections may merge in the camera’s imager and ruin the capture data. Merged marker reflections will have an inaccurate centroid location, or they may even be completely ignored by the circularity filter. We have a variety of different methods for discarding or salvaging the data. However, for best results, such situation should be prevented in the first place. Especially for precise tracking applications, marker placements and camera placements should be optimized in order to avoid the intrusion.

MVQSG Optimization4.png