Multiple Object Tracking using Fuzzy Logic for Handling Uncertainty

This paper proposes a new tracking method that is more suitable than traditional methods. The target model of traditional trackers based on the maximum a posteriori estimation (MAP) has high uncertainty because of various disturbance factors. To solve these problems, we designed a new tracker generated by a single sensor by using the minimum uncertainty gap (MUG) estimation, which considers not only the maximum average likelihood score, but also the minimum gap between the lower and upper bounds of likelihood. To simultaneously consider two likelihoods, we assign the weights inferred by using a fuzzy logic. At this point, the fuzzy inference system makes adaptive weights that can be assigned without any distortion modeling. The fuzzy logic is constructed by incorporating expert knowledge to adaptively allocate the weights of each likelihood. Our method is evaluated using public datasets. The results of our evaluation show that our proposed method has a higher accuracy than previous methods in tracking multiple moving objects.