Multimodal Fusion-Based Detection for Small-Scale Underwater Target Tracking Using Acoustic–Optical Imaging

Low underwater optical visibility and weak acoustic feature pose significant challenges for small-targets detection, such as suspended cables. To overcome these issues, this study proposes an acoustic–optical fusion framework for underwater small-scale target detection, by incorporating result-based filtering through interactive multiple model adaptive Kalman filter for dynamic tracking. First, a supervised enhancement network is developed with edge-preserving architecture to resolve feature blurring in low-quality acquisitions. Moreover, a result-based filtering framework processes detection outputs from acoustic–optical networks through probabilistic confidence allocation based on motion continuity characteristics, explicitly addressing optical detection uncertainties by fusing multisource sensor results. Underwater field experiments were conducted by using a typical mechatronic system—an uncrewed underwater vehicle equipped with both acoustic and optical sensors to detect and track a suspended cable. The results demonstrate that, in addition to achieving high tracking performance, the proposed method exhibits significantly stronger robustness and reliability compared to conventional approaches, and the proposed framework can be widely applied to synchronize and fuse multirate acoustic and optical data in underwater perception systems.