Artificial reef detection method for multi-beam sonar imagery by combining residual module with SegNet
With respect to how to automatically and accurately extract artificial reef areas from multi-beam sonar images,an artificial reef detection method for multi-beam sonar images by combining residual module with SegNet is proposed in this paper.The residual module can optimize the training network parameters to prevent the gradient dispersion and explosion in network training.The SegNet framework has the advantage of symmetry in the structure of the network encoding stage and decoding stage,and can achieve proportional symmetry between the network input image and output classification result.In this paper,the artificial reef detection in multi-beam sonar images is realized by integrating the residual module optimization network parameters and the symmetrical advantage of the SegNet structure.By testing different types of multi-beam artificial reef images,the proposed method can obtain 86.85%F1-socre and 76.75%IOU detection results,and accurately detect the artificial reef areas in the images.
万方数据
维普
