Fusion Tracking Algorithm of Mean-shift and Particle Filter Based on EMD

Real-time and accuracy are key indicators of object tracking. Traditional particle filter tracking requires calculating a large number of particles, which influences its real-time performance, Mean-shift is a non-parametric kernel density iterative algorithm, which is easily prone to local optimum and converges on non-real target. This paper proposes a new fusion tracking algorithm. It firstly extracts HOG features from the target, calculates the similarity between the target and candidates, and constructs the likelihood function. For different environments, it uses particle filter or Mean-shift algorithm. Experiments show that, it can still ensure real-time and accuracy even in the complex environment.