Multi Sensor Signal-based Multi-level Association strategy for Real-time Track Association and Fusion

The real-time track association and fusion task(the AF task) aims to integrate the tracks of maritime targets collected by sensors into a coherent track. In real-world scenarios, track data collected by a single sensor often lacks the required stability, falling short of meeting the demands of various applications. While adopting multi-sensor-based AF task methods is prevalent, existing approaches have overlooked critical aspects, including temporal consistency. This paper tackles crucial challenges in the multi-sensor AF tasks, focusing on problems such as asynchronous radar scanning and real-time matching across multiple sensors. Utilizing multi-level association and integrated learning algorithms, we enhance matching accuracy and efficiency while ensuring track continuity. Leveraging the dataset provided by the 2023 National Big Data and Computational Intelligence Challenge
¹
, our approach significantly outperforms traditional methods in three key aspects: the AF task of radar point trace , the AF task of ESM bearing line, and the AF task of rader and ESM. In comparison with conventional methods, three values are reduced by 1.1325, 0.0870, and 0.2637 on the difference metrics between the generated tracks and the actual tracks, respectively. These results further validate the significant improvements of our method in enhancing the effectiveness of the AF task.