Feature Tracks are not Zero-Mean Gaussian
This challenges a foundational assumption in state estimation for robotics and computer vision, potentially requiring algorithm redesigns.
The paper tackled the problem of inaccurate state estimation in algorithms by showing that feature track errors are not zero-mean Gaussian, as commonly assumed, and found that error distributions depend on motion type, speed, and image processing algorithms.
In state estimation algorithms that use feature tracks as input, it is customary to assume that the errors in feature track positions are zero-mean Gaussian. Using a combination of calibrated camera intrinsics, ground-truth camera pose, and depth images, it is possible to compute ground-truth positions for feature tracks extracted using an image processing algorithm. We find that feature track errors are not zero-mean Gaussian and that the distribution of errors is conditional on the type of motion, the speed of motion, and the image processing algorithm used to extract the tracks.