An identification algorithm for typical marine disaster-bearing bodies under small-sample conditions

The individual identification and refined management of marine disaster-bearing bodies are of great significance for accurate warning and evaluation of marine disasters. Due to the difficulty and high cost of obtaining a large amount of sample data for marine disaster-bearing bodies, this article conducts research on the identification method of marine disaster-bearing bodies under small-sample conditions for several typical marine disaster-bearing bodies. This article introduces the attention mechanism SKNet with multi-scale adaptive ability based on the ResNet-101 network, constructs a new feature extraction network, and uses this network as the backbone feature extraction network to construct a dual path twin neural network structure, which is used to train small-sample datasets and obtain a recognition model for marine disaster-bearing bodies. According to the experimental data, compared with the original ResNet-101 network, the improved network enhances the network’s ability to extract image features of complex scenes under the condition of small-sample, and has higher recognition accuracy and stability for marine disaster-bearing bodies.