A robust PHD filter with deep learning updating for multiple human tracking

Abstract

We propose a novel robust probability hypothesis density (PHD) filter for multiple target tracking in an enclosed environment, where a deep learning method is used in the update step for combining different human features to mitigate the effect of measurement noise on the calculation of particle weights. Deep belief networks (DBNs) are trained based on both colour and oriented gradient (HOG) histogram features and then used to mitigate the measurement noise from the particle selection and PHD update step, thereby improving the tracking performance. To evaluate the proposed PHD filter, two sequences with 383 frames from the CAVIAR dataset are employed and both the optimal subpattern assignment (OSPA) and mean of error from each target method are used as objective measures. The results show that the proposed robust PHD filter outperforms the traditional PHD filter.