Autonomous Underwater Vehicle Navigation Using Sonar Image Matching based on Convolutional Neural Network. Issue 21 (2019)

• Record Type: Journal Article
• Authors: Yang, Wenli; Fan, Shuangshuang; Xu, Shuxiang; King, Peter; Kang, Byeong; Kim, Eonjoo
• Is Part Of: IFAC-PapersOnLine. Volume 52:Issue 21(2019)
• Page Start: 156
• Page End: 162
• Publication Date: 2019
• Subjects: Sonar Image matching -- Convolutional neural network -- feature extraction -- AUV -- Teach -- Repeat path following; Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2019.12.300
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: This paper presents an image matching algorithm based on convolutional neural network (CNN) to aid in the navigating of an Autonomous Underwater Vehicle (AUV) where external navigation aids are not available. We aim to solve the problem where traditional image feature representations and similarity learning are not learned jointly and to improve the matching accuracy of sonar images in deep ocean with dynamic backgrounds, low-intensity and high-noise scenes. In our work, the proposed CNN-based model can train the texture features of sonar images without any manually designed feature descriptors, which can jointly optimize the representation of the input data conditioned on the similarity measure being used. The validation studies show the feasibility and veracity of the proposed method for many general and offset cases using collected sonar images.