Density-Sensitive Gaussian Mixture Model for Multi-Object Data Association

The establishment of data association is a vital component to ensure the real-time and precision of object tracking. In view of the deficiency of real-time and precision of traditional data association algorithms, this paper proposes a new data association model named density-sensitive Gaussian mixture model (DSGMM) to achieve multi-object data association tasks. To be specific, the multi-target association is first formulated as a clustering problem based on the Gaussian mixture model. Then, the proposed method establishes bidirectional constraint and confirmation matrix constraint criteria to correct the interconnection probability of the association model, aiming to improve the accuracy and efficiency of data association. Further, the idea of relative density is introduced to enhance the association probability of object points and weaken the association probability of clutter and noise, thereby improving the association accuracy. A series of simulation experiments illustrate that the proposed method is superior to some state-of-the-art data association methods.