Camera Selection Guide

The choice of cameras for your setup is critical for the size of the capture volume and accuracy of your motion capture system. This section explains how each camera spec affects the capture system.

Camera specs

Shutter type

重要

Unless you know what you are doing, select camera sensors that have a global shutter.

Rolling shutter generally produces distortion or blur when subjects make quick moves, making it difficult to track keypoints.

Sensor resolution

A large resolution can greatly benefit tracking performance. A subject needs to take up a certain amount of pixels to be practically trackable in the system, with more pixels increasing accuracy. How many pixels a subject takes up also depends on the camera's field of view.

Lens focal length (field of view)

The field of view (FOV) of the camera is another critical parameter. It is determined by the focal length of your lens, together with the width and height of the camera sensor.

  • Larger FOV: the capture volume will be larger (see Camera distribution).

  • Trade-off: objects at a fixed distance take up fewer pixels in the captured images, which reduces the effective range of the camera.

Tip

  • Body tracking: tracked subjects at the furthest distance should be at least 100 px long in the captured images.

  • Finger tracking: hands must be close enough to be around 100 px long as well.

注釈

Avoid cameras with extreme amounts of distortion such as fisheye lens. They are difficult to calibrate their intrinsics (see Intrinsic calibration).

Framerate

It is usually easy to find cameras with an acceptable framerate for motion capture.

Framerate

Suitability

30 FPS

May be acceptable, depending on your goals

60 FPS

Recommended

> 60 FPS

Little additional advantage

Serial Number uniqueness

To tell cameras apart across sessions, Kinoca identifies each camera by its serial number, which should be unique to every device. Unfortunately, some manufacturers of cheap cameras ignore this field or assign the same value to all produced units. In that case, Kinoca falls back to identifying cameras by the USB port they are plugged into, so each camera must stay on the same port between sessions (moving it to a different port requires updating the scene). Manufacturers rarely state whether their cameras have unique serial numbers, so if you are unsure, contact them before buying (see also example hardware).

Camera distribution

Capture volume size

At minimum, a 3D point in your location is part of the motion capture volume if it is visible by at least 2 cameras. For reliability against occlusions 3 or more cameras may be desirable.

注釈

You can always add more cameras later. Beginners can start with a few cameras to get a feel for the tracking capabilities, then upgrade.

Camera placement

Besides the volume size, cameras should be placed to minimize occlusions. An occlusion happens when certain keypoints of a body are hidden because other body parts or objects are in front of them. To reduce occlusions:

  • Point your cameras in diverse directions.

  • Mount them at differing heights around your location.