A Dual-Path Deep Neural Network for Sonar Image Quality Evaluation

Sonar technology plays an important role in the development of marine resources and military strategy. Due to the bad underwater acoustic channel, the sonar image collected by sonar technology equipment is affected by various kinds of distortions easily. To obtain high-quality sonar image, we devise a novel dual-path deep neural network (DPDNN) to measure the quality of sonar image. In these two paths, we use the batch normalization layer to reduce the training time and take the skip operation to speed up the feature extraction. Based on the above two operations, we extract the micro-scopic and macro-scopic structure of sonar image, respectively. Finally, the global average pooling layer and the fully connection layer are used to connect the above two paths. Experiments show that our DPDNN has a significant improvement in prediction performance and efficiency, respectively.