Parallel Fusion of Fuzzy Logic and Neural Network for Multisource Trajectory Association

Trajectory association, which forms precise trajectories of multitarget, is a pivotal technology that enhances the anti-interference capability of tracking by integrating data from diverse sources. It confronts challenges including inconsistent measurement error from distinct sensors, decreasing accuracy brought out by increasing target numbers, etc. In this work, we propose an innovative approach that combines fuzzy logic and neural network (FLNN) technology to achieve multitarget association. Specifically, statistical and instantaneous features are extracted from the trajectory data at first, and then a fuzzy logic system in parallel with a deep learning model is proposed to process these features. Afterward, the parallel results are integrated via evidence theory and filtered through the maximum association probability search (MAPS) algorithm, achieving trajectory association. Through validation in various simulation scenarios, the average ${F}1$ -score of the method reaches 95.1%, which is 7.6%–6.2% higher than the independent fuzzy logic module and the independent neural network module, respectively. Additionally, the testing with real-world data indicates that the proposed method outperforms several other representative association methods, which is consistent with the simulation experiments, proving its applicability in the real world.