Underwater Acoustic Target Classification and Grading Technology Based on Track Fusion

Aiming at the prominent problem of low accuracy of underwater acoustic target recognition based on acoustic features, we propose to fuse and use non-acoustic information to improve the confidence of target recognition. In this paper, an underwater acoustic target recognition technology based on acoustic and non-acoustic information fusion is proposed to tackle the problems of difficult track association and insufficient fusion probability model support in the process of acoustic and non-acoustic information fusion. This technology uses the improved target-tracking grey relational analysis (GRA) method to associate the tracks of low-frequency sonar targets, high-frequency sonar targets, and non-acoustic targets. The track information after association is fused based on Dempster-Shafer (D-S) evidence theory to give the non-threat assessment results of targets. Finally, the expert system is used to fuse the recognition results based on the line spectrum feature template matching of target radiated noise with the target non-threat assessment results based on acoustic and non-acoustic track fusion, so as to improve the accuracy of target recognition. The simulation results of the improved target-tracking GRA is shown and the feasibility of the proposed method is verified by sea trial data.