Deep Learning Convolutional Neural Network Applying for the Arctic Acoustic Tomography Current Inversion Accuracy Improvement

Warm current has a strong impact on the melting of sea ice, so clarifying features plays very important role in Arctic ice coverage forecasting study field. Currently, acoustic tomography is only feasible method for large-range measurement under ice. Furthermore, affected by high latitudes Coriolis force, small-scale variability greatly affects accuracy tomography. However, could not be measured empirical parameters and resolved Regularized Least Squares (RLS) inverse problem In this paper, convolutional neural network (CNN) proposed to enhance prediction Arctic, especially, Gaussian noise added reflect disturbance environment. First, we use finite element build background ocean model. Then, deep learning CNN constructs non-linear mapping relationship between data corresponding flow velocity. Finally, simulation result shows that being applied achieve 45.87% accurate improvement than common RLS inversion.