Track Segment Association Based on LSTM Networks

In target tracking under complex radar environments, many factors may lead to the interruption of the target trajectory, such as target maneuvering, low target detection probability, and clutter masking by sea or ground clutter. Existing algorithms for Track Segment Association (TSA) face challenges such as high computation complexity, high demand for prior information on targets, and weak ability to associate multiple target trajectories, making it difficult to cope with the complex and variable radar scenarios. Therefore, this paper proposes a multiple-target TSA model based on Long Short-Term Memory (LSTM) networks. The algorithm composes a feature matrix of the characteristics of interrupted trajectories before and after correlation. It utilizes an LSTM sequence-to-label model to solve the classification problem of whether the new and old trajectories are associated. Experimental results on measured data show that the proposed LSTM-based TSA method can effectively improve the continuity and accuracy of target tracking.