Difference between revisions of "Camera Placement"

Line 18: Line 18:
 
To calculate 3D marker locations, tracked markers must be simultaneously captured by at least two (or more depending on the [[Reconstruction]] settings.) cameras in the system. When a marker is occluded and not enough cameras are capturing the 2D positions, the 3D marker will not be present in the captured data. As a result, the collected marker trajectory will have gaps, and the accuracy of the capture will be reduced. Furthermore, extra effort and time will be required for [[Data Editing|post-processing]] the capture data. Thus, marker visibility throughout the capture is very important for tracking quality, and cameras need to be capturing at diverse vantages so that marker occlusions are minimized.
 
To calculate 3D marker locations, tracked markers must be simultaneously captured by at least two (or more depending on the [[Reconstruction]] settings.) cameras in the system. When a marker is occluded and not enough cameras are capturing the 2D positions, the 3D marker will not be present in the captured data. As a result, the collected marker trajectory will have gaps, and the accuracy of the capture will be reduced. Furthermore, extra effort and time will be required for [[Data Editing|post-processing]] the capture data. Thus, marker visibility throughout the capture is very important for tracking quality, and cameras need to be capturing at diverse vantages so that marker occlusions are minimized.
  
Depending on captured motion types and volume settings, the instructions for ideal camera arrangement may vary. For example, applications that require tracking markers at low heights, it would be beneficial to have some cameras placed and aimed at low elevations. For applications tracking markers placed on the front of the subject, cameras on the rear will be less beneficial because those markers will not be visible from behind. For large volume setups, installing cameras circumnavigating the volume at the highest elevation will maximize the use of camera coverages and the capture volume size. For captures valuing the accuracy, it is better to place cameras close to the object, so that cameras capture bigger images and track small changes in their position.  
+
Depending on captured motion types and volume settings, the instructions for ideal camera arrangement vary. For applications that require tracking markers at low heights, it would be beneficial to have some cameras placed and aimed at low elevations. For applications tracking markers placed strictly on the front of the subject, cameras on the rear won't see those and as a result become unnecessary. For large volume setups, installing cameras circumnavigating the volume at the highest elevation will maximize camera coverage and the capture volume size. For captures valuing extreme accuracy, it is better to place cameras close to the object so that cameras capture more pixels per marker and more accurately track small changes in their position.  
  
 
{{Tip|Again, the optimal camera arrangement depends on the purpose and features of the capture application. Plan the camera placement specific to the capture application so that the capability of the provided system is fully utilized.  Please [http://optitrack.com/contact/ contact us] if you need consulting with figuring out the optimal camera arrangement.}}
 
{{Tip|Again, the optimal camera arrangement depends on the purpose and features of the capture application. Plan the camera placement specific to the capture application so that the capability of the provided system is fully utilized.  Please [http://optitrack.com/contact/ contact us] if you need consulting with figuring out the optimal camera arrangement.}}
Line 24: Line 24:
 
==General Guide==
 
==General Guide==
  
For common application tracking 3D positions and orientations of skeletons and rigid bodies, place the cameras on the periphery of the capture volume. This setup typically maximizes the camera overlap and minimizes wasted camera coverage. General tips include the following:
+
For common applications of tracking 3D position and orientation of skeletons and rigid bodies, place the cameras on the periphery of the capture volume. This setup typically maximizes the camera overlap and minimizes wasted camera coverage. General tips include the following:
  
* Mount cameras at the desired maximum height of the capture volume, and aim them so they look straight across from the top of the volume.
+
* Mount cameras at the desired maximum height of the capture volume.
  
 
* Distribute the cameras equidistantly around the setup area.
 
* Distribute the cameras equidistantly around the setup area.
  
* Aim cameras towards the target volume. For bigger sized volumes, aim few cameras towards the edge of the volume to provide better coverage along the side as well.
+
* Adjust angle of cameras and aim them towards the target volume.
  
* For cameras with rectangular FOVs, mount the cameras so the imager is in landscape mode. In very small setup areas, cameras can be aimed in portrait to increase vertical coverage, but this typically reduces camera overlap, which can reduce marker continuity and data quality.
+
* For cameras with rectangular FOVs, mount the cameras in landscape orientation. In very small setup areas, cameras can be aimed in portrait orientation to increase vertical coverage, but this typically reduces camera overlap, which can reduce marker continuity and data quality.
  
 
{{Tip|'''TIP:''' For capture setups involving large camera counts, it is useful to separate the capture volume into two or more sections. This reduces amount of computation load for the software.}}
 
{{Tip|'''TIP:''' For capture setups involving large camera counts, it is useful to separate the capture volume into two or more sections. This reduces amount of computation load for the software.}}
Line 39: Line 39:
  
 
====Around the volume====
 
====Around the volume====
For common applications tracking a skeleton or a rigid body to obtain the 6 Degree of Freedom, (x,y,z-position and orientation) data, it is beneficial to arrange the cameras around the periphery of the capture volume for tracking markers both in front and back of the subject.
+
For common applications tracking a skeleton or a rigid body to obtain the 6 Degrees of Freedom (x,y,z-position and orientation) data, it is beneficial to arrange the cameras around the periphery of the capture volume for tracking markers both in front and back of the subject.
  
  
 
====Camera Elevations====
 
====Camera Elevations====
Placing cameras at high elevations is recommended because not only it maximize the capture coverage, but it also prevents subjects bumping into the setup, which could degrade the calibration. Furthermore, when cameras are placed at low elevations and aimed across from one another, the synchronized IR illuminations from each camera will be detected, and they will need to be [[Calibration#Masking|masked]] from the 2D view.  
+
Placing cameras at high elevations is recommended not only because it maximizes the capture coverage, but doing so also minimizes the chance of subjects bumping into the truss structure, which can degrade the calibration. Furthermore, when cameras are placed at low elevations and aimed across from one another, the synchronized IR illuminations from each camera will be detected, and will need to be [[Calibration#Masking|masked]] from the 2D view.  
  
  
However, it is also beneficial to place some cameras at varying elevations. It will provide more diverse view angles from both high and low elevations. If there are not enough cameras to cover all of the varying elevations, placing even one or more cameras at lower height will significantly improve the tracking. As a result, the frequency of marker occlusions will be reduced, and the accuracy for detecting the marker elevations will be improved.
+
However, it can be beneficial to place cameras at varying elevations. Doing so will provide more diverse viewing angles from both high and low elevations, and can significantly the coverage of the volume. The frequency of marker occlusions will be reduced, and the accuracy for detecting the marker elevations will be improved.
  
  
Line 65: Line 65:
 
====Camera to Camera Distance====
 
====Camera to Camera Distance====
  
Separating every camera at consistent distance is recommended. At least, there should be an established minimum separation between every camera to ensure that they are not placed too close to one another. When cameras are placed in close vicinity, they capture very similar images on the tracked subject, and the extra image will not contribute to preventing occlusions nor the reconstruction calculations. Instead, such setup will lose the benefit from the camera count, and also double the computation load for the calibration process. Moreover, this also increases the chance of marker occlusions because markers will be blocked from each view simultaneously whenever obstacles are introduced.
+
Separating every camera by a consistent distance is recommended. When cameras are placed in close vicinity, they capture similar images on the tracked subject, and the extra image will not contribute to preventing occlusions nor the reconstruction calculations. This overlap detracts from the benefit of a higher camera count, and also doubles the computational load for the calibration process. Moreover, this also increases the chance of marker occlusions because markers will be blocked from multiple views simultaneously whenever obstacles are introduced.
  
  
 
====Camera to Object Distance====
 
====Camera to Object Distance====
  
An ideal distance between camera and the capture subject also depends on the purpose of the capture. Allowing a long distance between the camera and the object gives more camera coverage for larger volume setups. On the other hand, capturing at a short distance will have less camera coverage but the tracking measurements will be more accurate. The camera focus may need to be adjusted for close-up tracking applications.
+
An ideal distance between camera and the capture subject also depends on the purpose of the capture. A long distance between the camera and the object gives more camera coverage for larger volume setups. On the other hand, capturing at a short distance will have less camera coverage but the tracking measurements will be more accurate. The cameras lens focus ring may need to be adjusted for close-up tracking applications.
  
  

Revision as of 18:35, 17 February 2016

Back to the Main PageBack to Hardware Setup


In optical motion capture systems, proper camera placement is very important in order to efficiently utilize the captured images from each camera. A well-arranged camera placement can significantly improve the tracking quality. When tracking markers, 3D coordinates are reconstructed from the 2D views seen by each camera in the system. More specifically, correlated 2D marker positions are triangulated to compute the 3D position of each marker. Thus, having multiple distinct vantages on the target volume is beneficial because it allows wider angles for the triangulation algorithm, which in turn improves the tracking quality. Accordingly, an efficient camera arrangement should have cameras distributed appropriately around the capture volume. By doing so, not only the tracking accuracy will be improved, but uncorrelated rays and marker occlusions will also be prevented.


Planning Camera Placement

An ideal camera placement varies depending on the capture application. In order to figure out the best placements for a specific application, a clear understanding of the fundamentals of optical motion capture is necessary.

To calculate 3D marker locations, tracked markers must be simultaneously captured by at least two (or more depending on the Reconstruction settings.) cameras in the system. When a marker is occluded and not enough cameras are capturing the 2D positions, the 3D marker will not be present in the captured data. As a result, the collected marker trajectory will have gaps, and the accuracy of the capture will be reduced. Furthermore, extra effort and time will be required for post-processing the capture data. Thus, marker visibility throughout the capture is very important for tracking quality, and cameras need to be capturing at diverse vantages so that marker occlusions are minimized.

Depending on captured motion types and volume settings, the instructions for ideal camera arrangement vary. For applications that require tracking markers at low heights, it would be beneficial to have some cameras placed and aimed at low elevations. For applications tracking markers placed strictly on the front of the subject, cameras on the rear won't see those and as a result become unnecessary. For large volume setups, installing cameras circumnavigating the volume at the highest elevation will maximize camera coverage and the capture volume size. For captures valuing extreme accuracy, it is better to place cameras close to the object so that cameras capture more pixels per marker and more accurately track small changes in their position.

Again, the optimal camera arrangement depends on the purpose and features of the capture application. Plan the camera placement specific to the capture application so that the capability of the provided system is fully utilized. Please contact us if you need consulting with figuring out the optimal camera arrangement.


General Guide

For common applications of tracking 3D position and orientation of skeletons and rigid bodies, place the cameras on the periphery of the capture volume. This setup typically maximizes the camera overlap and minimizes wasted camera coverage. General tips include the following:

  • Mount cameras at the desired maximum height of the capture volume.
  • Distribute the cameras equidistantly around the setup area.
  • Adjust angle of cameras and aim them towards the target volume.
  • For cameras with rectangular FOVs, mount the cameras in landscape orientation. In very small setup areas, cameras can be aimed in portrait orientation to increase vertical coverage, but this typically reduces camera overlap, which can reduce marker continuity and data quality.

TIP: For capture setups involving large camera counts, it is useful to separate the capture volume into two or more sections. This reduces amount of computation load for the software.


Camera Placement Checkpoints

Around the volume

For common applications tracking a skeleton or a rigid body to obtain the 6 Degrees of Freedom (x,y,z-position and orientation) data, it is beneficial to arrange the cameras around the periphery of the capture volume for tracking markers both in front and back of the subject.


Camera Elevations

Placing cameras at high elevations is recommended not only because it maximizes the capture coverage, but doing so also minimizes the chance of subjects bumping into the truss structure, which can degrade the calibration. Furthermore, when cameras are placed at low elevations and aimed across from one another, the synchronized IR illuminations from each camera will be detected, and will need to be masked from the 2D view.


However, it can be beneficial to place cameras at varying elevations. Doing so will provide more diverse viewing angles from both high and low elevations, and can significantly the coverage of the volume. The frequency of marker occlusions will be reduced, and the accuracy for detecting the marker elevations will be improved.


Cameras placed at periodical elevation

Placement Periodic.png


Cameras placed at random elevations

Placement Random.png


Cameras placed at alternating elevations

Placement Alternating.png

Camera to Camera Distance

Separating every camera by a consistent distance is recommended. When cameras are placed in close vicinity, they capture similar images on the tracked subject, and the extra image will not contribute to preventing occlusions nor the reconstruction calculations. This overlap detracts from the benefit of a higher camera count, and also doubles the computational load for the calibration process. Moreover, this also increases the chance of marker occlusions because markers will be blocked from multiple views simultaneously whenever obstacles are introduced.


Camera to Object Distance

An ideal distance between camera and the capture subject also depends on the purpose of the capture. A long distance between the camera and the object gives more camera coverage for larger volume setups. On the other hand, capturing at a short distance will have less camera coverage but the tracking measurements will be more accurate. The cameras lens focus ring may need to be adjusted for close-up tracking applications.


Examples

CameraPlacement Truss1.jpg CameraPlacement Tripod1.jpg CameraPlacement Truss2.jpg CameraPlacement Tripod2.jpg CameraPlacement Ondesk.jpg



Back: Hardware Setup

Next: Camera Mount Structures