A marine environment data fusion method and system based on attention mechanism

Abstract

The invention belongs to the technical field of data processing, and discloses a marine environment data fusion method and system based on an attention mechanism. The system includes a multi-source data enhancement module for constructing a time-space continuous data sequence and a The multi-layer feature combination neural network of the attention mechanism, the multi-layer feature combination neural network based on the attention mechanism includes an initial feature extraction layer, a deep feature interaction part and a fusion reconstruction layer, and the deep feature interaction part includes N groups of multi-layer feature combinations module, each multi-layer feature combination module includes M residual units, feature splicing layer, fusion convolution layer and channel attention module, the multi-layer feature combination module combines the features of different layers through global skip connection and local skip connection Information, the input of each multi-layer feature combination module is the output of the previous module. Through the fusion of feature information and attention mechanism in the present invention, the key information of features is highlighted.

Description

technical field [0001] The invention belongs to the technical field of data processing, and in particular relates to a marine environment data fusion method and system based on an attention mechanism. Background technique [0002] Time-space continuous marine environmental monitoring observation data is the basis for understanding the ocean. Accurate ocean observation data can be obtained through remote sensing observation technology, but the data is relatively discrete and cannot meet the needs of practical applications in the ocean field. How to obtain continuous marine monitoring data in time and space through observation data, that is, the fusion method of marine environmental data, is a key step in the utilization of marine environmental monitoring and observation data. [0003] The early researchers mainly adopted the physical model-driven scheme. This kind of method can achieve better results by parameterizing the mathematical and physical equations established for t...

AI Technical Summary

Problems solved by technology

The characteristics of the above data fusion tasks lead to the following shortcomings in the current deep learning-based marine environmental data fusion method: 1) Due to the characteristics of the satellite itself and the sensitivity to sampling conditions, the observation data is very sparse and may contain noise, which makes the model learning difficult to map

Therefore, only relying on deeper networks or deep features cannot better complete the task of ocean data reconstruction

3) The features extracted by each layer contain key information and supplementary information, and the model does not highlight the role of key information

Images