Multi Dimensional Feature Trajectory Association Network Based on Channel Attention Mechanism

This paper presents an innovative multi-dimensional feature trajectory correlation network leveraging a channel attention mechanism for multi-sensor trajectory correlation tasks. The model features distinct position and time extraction channels to mitigate the mutual influence between multidimensional features, capturing dynamic trends and periodic changes in trajectory data. To address the limitations of traditional methods with sparse trajectory point data, the model employs a channel attention mechanism to dynamically adjust feature channel weights, enhancing noise resistance and improving accuracy and stability in trajectory correlation. Additionally, a trajectory correlation detection dataset based on Beidou and Hainan fishing vessel AIS data was created for experimental evaluation. Comparative analysis reveals that the proposed method not only excels in accuracy and stability but also demonstrates superior adaptability and efficiency with large datasets and sparse trajectory points.