Underwater Sonar Image Segmentation Based on Deep Learning of Receptive Field Block and Search Attention Mechanism

Because of the special imaging environment, sonar images have some problems, such as gray distortion, blurred edge, various shapes, and missing dataset. To solve the missing of underwater sonar images dataset, a dataset of underwater sonar images dataset is established, including synthetic sonar dataset and real sonar dataset. According to the characteristics of sonar images, a new segmentation algorithm based on Deep CNN for Image Denoising (DnCNN) for image denoising is proposed, which integrates the Receptive Field Block and Attention Search Function. Experimental results show that the proposed method can effectively improve the accuracy of segmentation and retain more details of sonar images. At the same time, compared with other common segmentation methods, the proposed method also has significant advantages in network performance.