A comparison of two different tracking algorithms is provided for real application

The Multiple Hypothesis Tracker (MHT) and the Cardinalized Probability Hypothesis Density (CPHD) are two algorithms which can overcome the Multi-Targets Tracking (MTT) issues in automotive applications. This paper describes the performance of such algorithms and, in particular the Gaussian Mixture Probability Hypothesis Density (GMPHD) filter and the Track Oriented Multiple Hypothesis Tracker (TOMHT) for multiple cars and humans tracking in real road context. The scenario under consideration is the tracking an unknown number of real targets (humans and vehicles), using real measurements from an intelligent camera and a radar. The estimation of the number of targets and the target states of each filter will allow us to draw conclusion regarding the behavior of TOMHT and GMCPHD in real road context.