The Feature-level Fusion of AIS and Sonar Information based on LSTM-Transformer Joint Model

Abstract With the growth of global maritime trade and deeper exploration of marine re-sources, a large number of AIS signals are generated. The massive AIS data can provide enhanced support for underwater detection experiments. The fusion of AIS signals and sonar data can be utilized to improve the accuracy and efficiency of maritime surveillance and target identification. How to better match sonar and AIS data has become one of the key technological challenges. Traditional methods face several issues when dealing with sonar and AIS data, including data misalignment, difficulties in handling complex environmental noise, and poor anti-interference capabilities. To address these challenges, this study proposes a deep learning model called the LSTM-Transformer joint model, which utilizes LSTM for feature extraction followed by a multi-head attention mechanism to match sonar and AIS signals. This model can effectively fuse sonar and AIS signals at the feature level, significantly improving matching accuracy. Through comparing the performance of different models in processing data from multiple regions, this research found that the LSTM-Transformer joint model demonstrated the highest adaptability and accuracy in the experiments, where the R 2 value can reach 0.93 and the highest accuracy can be achieved in all experimental regions, proving its effectiveness and superiority in maritime surveillance and target identification tasks. Furthermore, even in the face of the challenge of temporal and spatial mismatch between AIS and sonar data, this model can still achieve efficient data fusion and accurate target identification.