Side-Scan Sonar Image Matching Method Based on Topology Representation

Dianyu Yang,Jingfeng Yu,Can Wang,Chensheng Cheng,Guang Pan,Xin Wen,Feihu Zhang

DOI: https://doi.org/10.3390/jmse12050782

IF: 2.744

2024-05-08

Journal of Marine Science and Engineering

Abstract:In the realm of underwater environment detection, achieving information matching stands as a pivotal step, forming an indispensable component for collaborative detection and research in areas such as distributed mapping. Nevertheless, the progress in studying the matching of underwater side-scan sonar images has been hindered by challenges including low image quality, intricate features, and susceptibility to distortion in commonly used side-scan sonar images. This article presents a comprehensive overview of the advancements in underwater sonar image processing. Building upon the novel SchemaNet image topological structure extraction model, we introduce a feature matching model grounded in side-scan sonar images. The proposed approach employs a semantic segmentation network as a teacher model to distill the DeiT model during training, extracting the attention matrix of intermediate layer outputs. This emulates SchemaNet's transformation method, enabling the acquisition of high-dimensional topological structure features from the image. Subsequently, utilizing a real side-scan sonar dataset and augmenting data, we formulate a matching dataset and train the model using a graph neural network. The resulting model demonstrates effective performance in side-scan sonar image matching tasks. These research findings bear significance for underwater detection and target recognition and can offer valuable insights and references for image processing in diverse domains.

oceanography,engineering, marine, ocean

What problem does this paper attempt to address?

This paper aims to address the challenges in side - scan sonar image matching. Specifically, side - scan sonar images play an important role in marine environment detection, but these images are usually of low quality, have complex features and are easily affected by distortion, which makes it very difficult to extract and match features from these images. Traditional feature - matching algorithms (such as SIFT, SURF) perform poorly when dealing with such images, and the bag - of - words - based methods are also difficult to be effectively applied in complex underwater environments.

To overcome these challenges, the author proposes a side - scan sonar image - matching method based on topological representation. The main contributions of this method are as follows:

1. **Knowledge Distillation**: Use a custom - made convolutional neural network (CNN) semantic segmentation model as a teacher model to train the DeiT model for knowledge distillation. By obtaining the output of the intermediate layer and the self - attention matrix, the DeiT model can learn a richer feature representation.

2. **Topological Structure Extraction**: Based on SchemaNet, use the clustering method to cluster the data of the intermediate layer into multiple visual words, which are used as the nodes of the image topological structure. At the same time, combine the self - attention matrix to calculate the weights of the nodes and their interaction coefficients to form the topological structure of the image.

3. **Graph Neural Network**: Use a graph convolutional network (GCN) to extract the feature vectors of the image topological structure and transform it into an image - matching problem. By training the graph neural network to make it output the topological similarity between different images, image matching is achieved.

In conclusion, this paper proposes an innovative method. By combining deep - learning and graph - neural - network technologies, it improves the accuracy and robustness of side - scan sonar image matching, which is of great significance for underwater detection and target recognition.