The increasing use of wireless iot devices and edge servers in maritime environments has sparked interest in distributed edge computing research. Federated Learning (FL) has become a key solution to address communication resource consumption and data privacy issues in distributed edge computing. Nevertheless, current FL frameworks might not fully consider resource limitations in challenging marine environments. In this paper, we introduce a novel FL algorithm for signal overlapping in marine scenarios, focusing on maritime communication environment modeling, Wireless Federated Learning Overlapping (FedOverlap) algorithm design, and resource allocation optimization. We create a detailed wireless signal propagation model for maritime environments and develop an iterative FL algorithm to tackle the challenge of slow model convergence due to signal overlapping. Moreover, we suggest a resource scheduling and allocation strategy for efficient bandwidth, energy, and computation usage. Comprehensive experiments validate our FedOverlap algorithm’s properties and exhibit superior performance in accuracy, resource utilization, and convergence speed for practical network parameters and benchmark datasets in production-ready settings.