A Multi-Target Track Association Algorithm in Underwater Multi-Sensors Environments

In order to solve the problem of multi-target trajectory correlation in underwater environments, such as the error of underwater measurement and the inconsistency of the targets reported by sensors, we leverage the gradation pre-processing to eliminate the noise and propose the Gaussian mixture model by the topology information between the trajectories. A novel mixed integer nonlinear programming is constructed to determine the relationship between underwater target trajectories, and the correlation bias is reduced by recursing the sensor bias estimation continuously. At the same time, the idea of weighting is introduced to maximize the clustering expectation. The optimal closure solution of the GMM is achieved with the expectation maximization clustering. The corresponding relationship of the trajectory is gotten at the expectation maximization stage, and finally obtains the trajectory correlation result of the underwater targets. The simulation results show that the GMP algorithm has better positive correlation rate and robustness than other algorithms with different target numbers in different dimensions, different sensor angular ranging errors, and different sensor detection probabilities. The GMP algorithm has advantages in the complex underwater environments with multiple noise and high false alarms, and it has accuracy and robustness in underwater multi-target trajectory correlation.