A Rapid Dehazing Model in USV Imaging System based on End-to-End Convolutional Network

Abstract

Due to the inclement weather at sea, the pictures collected by USV will always be affected by fog, degrading in quality. In order to improve the quality of images collected by USV, convolutional neural networks (CNN) are widely used in image defogging. In this article, we proposed a CNN-based end-to-end defogging neural network, using dense connection blocks and the attention mechanism to obtain image information. Since convolutional neural networks have the advantages of local perception and parameter sharing, we have changed the traditional apriori method for atmospheric scattering models. We use multi-scale learning to capture image information and use the attention mechanism to suppress some irrelevant details. Experiments have shown that our method has a good effect on USV image dehazing problem.