Lightweight Underwater Object Detection Algorithm for Embedded Deployment Using Higher-Order Information and Image Enhancement

Changhong Liu,Jiawen Wen,Jinshan Huang,Weiren Lin,Bochun Wu,Ning Xie,Tao Zou

DOI: https://doi.org/10.3390/jmse12030506

IF: 2.744

2024-03-19

Journal of Marine Science and Engineering

Abstract:Underwater object detection is crucial in marine exploration, presenting a challenging problem in computer vision due to factors like light attenuation, scattering, and background interference. Existing underwater object detection models face challenges such as low robustness, extensive computation of model parameters, and a high false detection rate. To address these challenges, this paper proposes a lightweight underwater object detection method integrating deep learning and image enhancement. Firstly, FUnIE-GAN is employed to perform data enhancement to restore the authentic colors of underwater images, and subsequently, the restored images are fed into an enhanced object detection network named YOLOv7-GN proposed in this paper. Secondly, a lightweight higher-order attention layer aggregation network (ACC3-ELAN) is designed to improve the fusion perception of higher-order features in the backbone network. Moreover, the head network is enhanced by leveraging the interaction of multi-scale higher-order information, additionally fusing higher-order semantic information from features at different scales. To further streamline the entire network, we also introduce the AC-ELAN-t module, which is derived from pruning based on ACC3-ELAN. Finally, the algorithm undergoes practical testing on a biomimetic sea flatworm underwater robot. The experimental results on the DUO dataset show that our proposed method improves the performance of object detection in underwater environments. It provides a valuable reference for realizing object detection in underwater embedded devices with great practical potential.

oceanography,engineering, marine, ocean

What problem does this paper attempt to address?

The paper mainly addresses the challenges faced in underwater object detection and proposes a lightweight underwater object detection algorithm. Specifically, the study tackles the following key issues:

1. **Image Quality Issues in Underwater Environments**: Due to factors such as refraction and scattering of seawater, underwater images often appear blurry and color-distorted. These issues pose significant challenges to vision-based underwater object detection.

2. **Limitations of Existing Models**: Existing underwater object detection models typically suffer from low robustness, high computational complexity of model parameters, and high false detection rates.

To address the above issues, the paper proposes the following solutions:

- **Image Enhancement and Data Augmentation**: First, FUnIE-GAN is used for color restoration of underwater images, and data augmentation techniques such as Mixup are employed to effectively capture object feature information.

- **Improved Object Detection Network**: An enhanced object detection network named YOLOv7-GN is designed, which includes a lightweight high-order attention layer aggregation network (ACC3-ELAN) to improve the fusion perception capability of high-order features in the backbone network. Additionally, the interaction of multi-scale high-order information further enhances the capability of the head network.

- **Lightweight Design**: To enable the deployment of the model on embedded devices, the AC-ELAN-t module is introduced, which is pruned based on ACC3-ELAN, helping to further reduce the number of parameters and computational complexity of the entire network.

In summary, this study aims to develop a lightweight and efficient underwater object detection method to overcome the inherent problems of object detection in underwater environments and provide technical support for practical application scenarios such as underwater robots.