Accuracy estimation of object tracking methods for identification of 2D-coordinates and velocities of mechanical systems based on digital photography data

Abstract

Nowadays, computer vision methods, in particular, object tracking, are widely used by scientists and engineers. Object tracking is primarily used in such applications as CCTV, crash tests, sports broadcasting, etc. Many methods can be adapted for non-contact optical measurements of coordinates and velocities of mechanical systems in physical experiments. The influence of camera parameters on the optical measurement accuracy is analyzed. We present a comparative review of tracking algorithms and show how they can be used in mechanical experiments and analyze their accuracy in real mechanical experiments. We also discuss our earlier developed methods adopted for the laboratory experiments, which have subpixel accuracy and are robust to lighting differences. We present a short review of tracking software. It is shown that object tracking techniques can be used for non-contact measurements and analysis of mechanical systems.