Symmetry Threshold

Abstract

The symmetry threshold dynamicSymmetryThreshold is used for filtering motion artifacts. Increasing the threshold value leads to more valid pixels around moving objects but also increases the chance of computing incorrect distance measurements for some pixels. Decreasing the threshold will result in invalidation of more pixels because of their estimated symmetry value. This filter uses the noise model to assess the likelihood of a pixel to be affected by motion.

Description

The O3R camera heads use the ifm ToF (Time of Flight) technology for measuring the distance to objects. To calculate one single point cloud image, the system takes several independent image frames. These images are correlated over time. This correlation is represented as symmetry value. It can be thought of as the four modulated signals used for performing the raw measurement, being more or less symmetrical to one another.

For low symmetry threshold values, only pixels for which the correlation images are highly symmetrical (that is, with no or few motion artifacts) are valid. Because of inherent noise, a perfect symmetry is never possible even for static scenes. Increasing the symmetry threshold validates pixels with higher symmetry values, including noisy pixels and potential motion artifacts.

The dynamic symmetry ensures that the symmetry threshold of a pixel is at least high enough to prevent invalidation due to sensor noise. It can be thought of as differentiating motion artifacts from ambient noise in the scene.