Multiple sperm tracking in microscopic videos using modified GM-PHD filter

Abstract

This paper presents a method for simultaneous tracking of multiple sperms using modified Gaussian mixture probability hypothesis density (GM-PHD) filter. In order to track sperms with spurious motion, a modified model is presented to adapt the GM-PHD filter for nonlinear dynamic movement of sperms. Furthermore, the “pruning” step in the GM-PHD filter is modified to handle situations like occlusion or closely moving targets. Our experiments demonstrate more effectivity of the proposed method in terms of sperms’ occlusion handling and trajectory extraction compared to the conventional GM-PHD filter. In particular, the new method performs well in managing the labels of occluded sperms after separation and in tracking of temporarily disappeared sperms when they emerge again in the tracking space.