Perception and Recognition of Underwater Cable Based on Deep Learning
Zhenyi Bi, Yu‐Long Wang, Lang Ma · 2020
Aiming at the problems of blurring, low contrast and high noise of underwater target data sets, this paper proposes an underwater image enhancement method based on weight fusion to enhance the image of underwater cables. This method introduces the idea of image fusion to fuse the pictures processed by the white balance algorithm and the auto level algorithm according to a certain weight. It achieves the performance of color correction and target enhancement. The characteristics of the underwater cable are clearer after the image enhancement preprocessing. In addition, an underwater target recognition system based on deep learning is established. The Faster RCNN and SSD deep learning algorithms are utilized to realize underwater cable recognition. Good recognition accuracy and performance are demonstrated through an experiment on data set of underwater cables.