Acoustic characterization of deep-sea manganese nodule coverage using multi-classifier decision fusion

Rich in multiple strategic rare metals, deep-sea manganese nodules constitute an essential mineral resource for balancing terrestrial resource supply and demand and facilitating the global energy transition. Accurate prediction of nodule coverage is a key parameter for evaluating deep-sea mineral resources. However, the complex and heterogeneous nature of the deep-sea environment means that traditional detection methods for surface sediments and manganese nodules primarily focus on categorical classification, making quantitative assessment of indicators such as seafloor polymetallic nodule coverage challenging. To address these challenges, this study proposes an acoustic characterization method for deep-sea manganese nodule coverage based on multi-classifier ensemble decision fusion. First, terrain and backscatter intensity features derived from deep-water multibeam sonar data were extracted and analyzed, and the Boruta algorithm was employed to optimize the high-dimensional feature space. Second, a robust sliding-window estimation method was applied to suppress gross errors in the feature images. Finally, an improved Stacking-based ensemble prediction model was developed to achieve continuous spatial characterization of nodule coverage through the integration of multiple classifiers. Validation using deep-sea multibeam sonar data and ground truth samples from the western Clarion–Clipperton Zone (CCZ) demonstrates that the proposed approach achieves a prediction accuracy of 93.21% and a Kappa coefficient of 0.8962. Compared with Random Forest, Support Vector Machine, Back Propagation Neural Network, Decision Tree, and K-Nearest Neighbors classifiers, the proposed model improved prediction accuracy by 2.35%, 39.06%, 14%, 7.33%, and 3.69%, respectively. The proposed method effectively delineates the spatial distribution of manganese nodule coverage and provides methodological support for environmental baseline assessment and resource exploration in deep-sea environments.