Self-Supervised Pre-Training Joint Framework: Assisting Lightweight Detection Network for Underwater Object Detection

Underwater image enhancement and object detection has great potential for studying underwater environments. It has been utilized in various domains, including image-based underwater monitoring and Autonomous Underwater Vehicle (AUV)-driven applications such as underwater terrain surveying. It has been observed that, underwater images are not clear due to several factors such as: low light, the presence of small particles, different levels of refraction of light etc. Extracting high-quality features from these images to detect objects is a significant challenging task. To mitigate this challenge, MIRNet and modified version of YOLOv3 namely Underwater-YOLOv3 (U-YOLOv3) is proposed. The MIRNetisadeeplearning-based technology for enhancing underwater images. while using YOLOv3 for underwater object detection it lacks in detection of very small objects and huge size objects. To address this problem proper anchor box size, quality feature aggregation technique, and during object classification image resizing is required. The proposed U-YOLOv3 has three unique features which helps to work with above specified issue like accurate anchor box determination using the K-means++ clustering algorithm, introduced Spatial Pyramid Pooling (SPP) layer during feature extraction which helps in feature aggregation and added downsampling and upsampling to improve detection rate of very large and very small size objects. The size of the anchor box is crucial in detecting objects of different sizes, SPP helps in aggregation of features, while down and upsampling changes sizes of objects during object detection. The experiment on the Brackish and Trash-ICRA19 datasets shows that our proposed method enhances the mean average precision for both datasets by 10% and 9%, respectively, compared to the original YOLOv3 and other existing work. This enhancement demonstrates that our proposed model is more appropriate for identifying submerged items and is also capable of recognizing closely clustered, small, very large objects on the ocean floor.

A challenging and attractive task in computer vision is underwater object detection. Although object detection techniques have achieved good performance in general datasets, problems of low visibility and color bias in the complex underwater environment have led to generally poor image quality; besides this, problems with small targets and target aggregation have led to less extractable information, which makes it difficult to achieve satisfactory results. In past research of underwater object detection based on deep learning, most studies have mainly focused on improving detection accuracy by using large networks; the problem of marine underwater lightweight object detection has rarely gotten attention, which has resulted in a large model size and slow detection speed; as such the application of object detection technologies under marine environments needs better real-time and lightweight performance. In view of this, a lightweight underwater object detection method based on the MobileNet v2, You Only Look Once (YOLO) v4 algorithm and attentional feature fusion has been proposed to address this problem, to produce a harmonious balance between accuracy and speediness for target detection in marine environments. In our work, a combination of MobileNet v2 and depth-wise separable convolution is proposed to reduce the number of model parameters and the size of the model. The Modified Attentional Feature Fusion (AFFM) module aims to better fuse semantic and scale-inconsistent features and to improve accuracy. Experiments indicate that the proposed method obtained a mean average precision (mAP) of 81.67% and 92.65% on the PASCAL VOC dataset and the brackish dataset, respectively, and reached a processing speed of 44.22 frame per second (FPS) on the brackish dataset. Moreover, the number of model parameters and the model size were compressed to 16.76% and 19.53% of YOLO v4, respectively, which achieved a good tradeoff between time and accuracy for underwater object detection.

Real-time underwater object detection technology for complex underwater environments based on deep learning

Due to medium scattering, absorption, and complex light interactions, capturing objects from the underwater environment has always been a difficult task. Single-pixel imaging (SPI) is an efficient imaging approach that can obtain spatial object information under low-light conditions. In this paper, we propose a single-pixel object inspection system for the underwater environment based on compressive sensing super-resolution convolutional neural network (CS-SRCNN). With the CS-SRCNN algorithm, image reconstruction can be achieved with 30% of the total pixels in the image. We also investigate the impact of compression ratios on underwater object SPI reconstruction performance. In addition, we analyzed the effect of peak signal to noise ratio (PSNR) and structural similarity index (SSIM) to determine the image quality of the reconstructed image. Our work is compared to the SPI system and SRCNN method to demonstrate its efficiency in capturing object results from an underwater environment. The PSNR and SSIM of the proposed method have increased to 35.44% and 73.07%, respectively. This work provides new insight into SPI applications and creates a better alternative for underwater optical object imaging to achieve good quality.

Region-specialized underwater image restoration in inhomogeneous optical environments

Underwater optical imaging is essential for exploring the underwater environment to provide information for planning and regulating underwater activities in various underwater applications, such as aquaculture farm observation, underwater topographical survey, and underwater infrastructure monitoring. Thus, there is a need to investigate the underwater imaging process and propose clear and long-range underwater optical imaging methods to fulfill the demands of academia and industry. In this manuscript, we classify the eighteen most commonly used underwater optical imaging methods into two groups regarding the imaging principle, (1) hardware and (2) software-based methods, each with an explanation of the theory, features, and applications. Furthermore, we also discuss the current challenges and future directions for improving the performance of current methods, such as improving the accuracy of underwater image formation model estimation, enlarging the underwater image dataset, proposing comprehensive underwater imaging evaluation metrics, estimating underwater depth and integrating different methods (e.g., hardware- and software-based methods for computational imaging) to promote the imaging performance not only in the laboratory but also in practical underwater scenarios.

In recent years, underwater object detection (UOD) has become a prominent research area in the computer vision community. However, existing UOD approaches are still vulnerable to underwater environments, which mainly include light scattering and color shifting. The blurring problem caused by water scattering on underwater images makes the high-frequency texture edge less obvious, affecting the detection effect of objects in the image. To address this issue, we design a multi-scale high-frequency information enhancement module to enhance the high frequency features extracted by the backbone network and improve the detection effect of the network on underwater objects. Another common issue caused by scattering and color shifting is that it can easily change the low-frequency information in the background of underwater images, leading to performance degradation of the same target in different underwater scenes. Therefore, we have also designed a multi-scale gated channel information optimization module to reduce the scattering and color shifting effects on the channel information of underwater images and adaptively compensate the features for different underwater scenes. We tested the detection performance of our designed method on three typical underwater object detection datasets, RUOD, UDD and UODD. The experimental results proved that our method performed better than existing detection methods on underwater object detection datasets.

In recent years, deep learning based methods have achieved excellent performance in standard object detection. However, these methods lack sufficient capabilities to handle underwater object detection due to these challenges: (1) underwater images have color shift and low contrast, and (2) various noises (sediments, bubbles, etc.) would cause blurring in underwater images. To address these problems, our work investigates optimization policies to detect low-contrast and blurred mine. We propose a bounding box repairing algorithm based on Intersection over Union (IoU) optimization. The algorithm mainly optimizes Mask Scoring R-CNN network. Firstly, we use the Mask Scoring R-CNN model to generate the coarse-grained detection results of mines. Secondly, we perform IoU between the coarse-grained boxes and the annotated boxes in the dataset to obtain the best matching area. Finally, we use the best matching area to repair the coarse-grained boxes. Experimental results show that our method achieves promising object detection and localization performance in underwater environment.

Improved YOLOv7 model for underwater sonar image object detection

Underwater images often suffer from low contrast, color distortion, and noise due to light absorption and scattering in water. These issues make it challenging for marine applications such as underwater exploration, object detection, and autonomous navigation. Traditional image enhancement techniques struggle to recover natural colors and details effectively. This research proposes a U-Net-based Generative Adversarial Network (U-Net GAN) to enhance underwater images, focusing on preserving structural details while improving visual quality. The U-Net architecture ensures effective feature extraction and segmentation, while the GAN framework enhances image realism by learning high-level transformations. The model is trained and tested on benchmark datasets such as UIEB and EUVP, with performance evaluated using PSNR (Peak Signal-to-Noise Ratio), SSIM (Structural Similarity Index), and UIQM (Underwater Image Quality Measure). Experimental results demonstrate that the proposed approach significantly enhances underwater images compared to conventional methods, restoring natural colors and improving visibility. This study contributes to advancing deep learning techniques for real-time underwater image enhancement, benefiting various marine research and surveillance applications.

Object detection in underwater environments presents significant challenges due to the inherent limitations of sonar imaging, such as noise, low resolution, lack of texture, and color information. This paper introduces AquaYOLO, an enhanced YOLOv8 version specifically designed to improve object detection accuracy in underwater sonar images. AquaYOLO replaces traditional convolutional layers with a residual block in the backbone network to enhance feature extraction. In addition, we introduce Dynamic Selection Aggregation Module (DSAM) and Context-Aware Feature Selection (CAFS) in the neck network. These modifications allow AquaYOLO to capture intricate details better and reduce feature redundancy, leading to improved performance in underwater object detection tasks. The model is evaluated on two standard underwater sonar datasets, UATD and Marine Debris, demonstrating superior accuracy and robustness compared to baseline models.

There have been increased developments in ocean exploration using autonomous underwater vehicles (AUVs) and unmanned underwater vehicles (UUVs). However, the contrast of underwater images is still a major issue for application. It is difficult to acquire clear underwater images around underwater vehicles. Since the 1960s, sonar sensors have been extensively used to detect and recognize objects in oceans. Due to the principles of acoustic imaging, sonar-imaged images have many shortcomings, such as a low signal to noise ratio and a low resolution. Consequently, vision sensors must be used for short-range identification because sonars yield to low-quality images. This thesis will concentrate solely on the optical imaging sensors for ocean observing. Although the underwater optical imaging technology makes a great progress, the recognition of underwater objects also remains a major issue in recent days. Different from the common images, underwater images suffer from poor visibility due to the medium scattering and light distortion. First of all, capturing images underwater are difficult, mostly due to attenuation caused by light. The random attenuation of the light mainly causes the haze appearance along with the part of the light scattered back from the water. In particular, the objects at a distance of more than 10 m are almost indistinguishable because of absorption. Furthermore, when the artificial light is employed, it can cause a distinctive footprint on the seafloor. In this paper, we will analysis the recent trends of ocean exploration approaches.

Underwater object detection faces various challenges like turbidity, non-uniform lighting, scattering, and lack of image clarity. The application of deep learning techniques for underwater object detection has significantly improved the detection performance as compared to the traditional object detection methods. However not many attempts have been made to detect trash from underwater images mostly due to lack of images and labeled ground truth datasets. This chapter aims to evaluate the YOLACT and Faster-RCNN algorithm to perform trash detection in an underwater environment. A freely available dataset of marine debris is used to train the CNN for object detection. The trained network is then evaluated on test images to assess its fitness for real-time applications.

PE-Transformer: Path enhanced transformer for improving underwater object detection

The performance of deep neural networks in object detection is impressive, however, direct use of existingalgorithms may not obtain satisfactory resultsdue to the characteristics of underwater images. In this paper, a multi-scale feature fusion Single Shot Multibox Detector (MFFSSD) based on the SSD algorithm is proposed to improve the accuracy of underwater object detection. There are two main improvements: introducing attention module and perform multiscale feature fusion. By introducing attention module, the features of the input image obtained through convolution are weighted in both channel and space aspect, so as to highlight the useful informationto overcomes the problem of small and blur objects which are common in underwater images. By extracting a lower level feature map forward, more details of the lower level image can be obtained to adapt to the underwater object with simple shape, so that low-level features have a higher contribution to the recognition object; at the same time, the high-level feature map is enlarged by using deconvolution, and then fused with the low-level feature map to introduce more contextual information. The fusion of the features extracted from the shallow and high level networks ensures the efficiency of the features to the greatest extent. Experiments on URPC show that MFFSSD significantly improved the underwater object recognitionaccuracy compared to the original algorithm.