Last Revision Date: 6/6/2014
This article describes the accuracy of our Vision disparity calculations, and provides some general suggestions on testing and verifying the quality of the disparity calculations.
The accuracy of disparity calculations are very dependent on what the camera is being pointed at, but it is generally 1 - 2mm at up to approximately 2 metres. For a brief summary of depth resolution over a variety of operating distances and fields of view, or to calculate approximate accuracy measurements based on camera options, download our interactive Stereo Accuracy Chart (Microsoft Excel format).
In terms of testing and verifying the quality of the disparity calculations, it is very difficult to do any absolute measurements - the reason for this is that the origin (0,0,0) of the camera is unknown and can vary from camera to camera.
The quality of disparity results will vary from environment to environment. However, determining the quality of disparity calculations in your environment could be done relatively in a couple of different ways:
- Point the camera at a feature, determine the position of the feature relative to the camera, then move the feature a known amount and re-determine its location. This should at least provide an idea of how good the distance calculation is.
- Point the camera at a highly textured planar surface, capture all of the points on the surface, fit a plane to the points and then determine how far off each pixel is from the plane. Again this will provide an idea of the relative accuracy of the camera.