Explore the vast range of titles available.

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.

Underwater target classification has been an important topic driven by its general applications. Convolutional neural network (CNN) has been shown to exhibit excellent performance on classifications especially in the field of image processing. However, when applying CNN and related deep learning models to underwater target classifications, the problems, including small sample size of underwater target and low complexity requirement, impose a great challenge. In this paper, we have proposed the modified DCGAN model to augment data for targets with small sample size. The data generated from the proposed model help to improve classification performance under imbalanced category conditions. Furthermore, we have proposed the S-ResNet model to obtain good classification accuracy while significantly reducing complexity of the model, and achieve a good tradeoff between classification accuracy and model complexity. The effectiveness of proposed models is verified through measured data from sea trial and lake tests.

This paper proposes a deep convolutional neural network (DCNN) to design an accurate active sonar image classifier. In order to have a real-time classifier with low complexity, The LeNet-5 is utilized as the most straightforward deep network with the fewest parameters. For the sake of having a real-time training and test phase, the three fully connected layers are replaced by an extreme learning machine (ELM). However, tuning the ELM’s input layer parameters is challenging; therefore, this paper tries to tune them using the grey wolf optimizer (GWO). Contrary to other research works and considering the sonar problem’s characteristics, we model the problem as a multimodal function. Therefore, comprehensive learning concepts and a novel constraint-handling technique are exerted on the GWO to address the multimodality and the constraints of the sonar image classification task and to have a robust optimizer. Given the vital role of the reliable dataset in deep learning approaches, in the following, an operational underwater sonar test scenario is designed, and an experimental dataset is generated. The designed model is then benchmarked on two benchmark active sonar datasets. The results are investigated by qualified research with classic DCNN, Block-wise Classifier (BWC), and Matched Subspace classifier with Adaptive Dictionaries (MSAD). The investigation outcomes confirm that the designed model, with an average accuracy of 98.69% and computation time equal to 883.44 s, reports the best accuracy and time complexity among other benchmark models.

Underwater acoustic classification plays a critical role in both commercial and military applications. It faces significant challenges due to high background noise and the dynamic nature of marine environments. Factors such as temperature, pressure, and salinity create complex sound propagation patterns, causing feature space variations. These variations cause degraded performance of classifications systems when deployed in unseen conditions. Traditional approaches often rely on isolated learning paradigms, limiting their adaptability across diverse sea environments. To overcome this, we propose a novel learning strategy that dynamically extracts, retains, and transfers knowledge to improve generalization. Our method introduces an adaptive regularization technique for stable weight and learning rate updates. In addition, a robust knowledge encoding scheme is introduced to preserve and reuse learned features effectively. To demonstrate the effectiveness of the proposed method, extensive experiments are conducted on benchmark underwater acoustic datasets collected from various environments. The experimental results show that the proposed method achieves superior performance in diverse sea environments.

Efficient deep learning models are crucial in resource-constrained environments, especially for marine image classification in underwater monitoring and biodiversity assessment. This paper presents LatentResNet, a computationally lightweight deep learning model involving two key innovations: (i) using the encoder from the proposed LiteAE, a lightweight autoencoder for image reconstruction, as input to the model to reduce the spatial dimension of the data and (ii) integrating a DeepResNet architecture with lightweight feature extraction components to refine encoder-extracted features. LiteAE demonstrated high-quality image reconstruction within a single training epoch. LatentResNet variants (large, medium, and small) are evaluated on ImageNet-1K to assess their efficiency against state-of-the-art models and on Fish4Knowledge for domain-specific performance. On ImageNet-1K, the large variant achieves 66.3% top-1 accuracy (1.7M parameters, 0.2 GFLOPs). The medium and small variants reach 60.8% (1M, 0.1 GFLOPs) and 54.8% (0.7M, 0.06 GFLOPs), respectively. After fine-tuning on Fish4Knowledge, the large, medium, and small variants achieve 99.7%, 99.8%, and 99.7%, respectively, outperforming the classification metrics of benchmark models trained on the same dataset, with up to 97.4% and 92.8% reductions in parameters and FLOPs, respectively. The results demonstrate LatentResNet’s effectiveness as a lightweight solution for real-world marine applications, offering accurate and lightweight underwater vision.

Underwater target classification methods based on deep learning suffer from obvious model overfitting and low recognition accuracy in the case of small samples and complex underwater environments. This paper proposes a novel classification network (EfficientNet-S) based on EfficientNet-V2S. After optimization with model scaling, EfficientNet-S significantly improves the recognition accuracy of the test set. As deep learning models typically require very large datasets to train millions of model parameter, the number of underwater target echo samples is far more insufficient. We propose a deep convolutional generative adversarial network (SGAN) based on the idea of group padding and even-size convolution kernel for high-quality data augmentation. The results of anechoic pool experiments show that our algorithm effectively suppresses the overfitting phenomenon, achieves the best recognition accuracy of 92.5%, and accurately classifies underwater targets based on active echo datasets with small samples.

Multispectral LiDAR (MSL) systems offer a significant advantage by actively capturing both spatial and spectral information. These systems offer significant promise in supporting the comprehensive analysis and precise classification of underwater targets. In this study, we build an MSL system based on an acousto-optic tunable filter (AOTF) to investigate the feasibility of underwater sphere classification. The MSL prototype features a spectral resolution of 20 nm and 13 spectral channels, covering a range from 560 to 800 nm. Laboratory-based experiments were conducted to evaluate the accuracy of range measurements and the classification performance of the system. The spectral curves of nine distinct spheres acquired by the MSL were utilized for classification using a support vector machine (SVM). The experimental results indicate that classification using multispectral data yields a higher accuracy and Kappa coefficient. Finally, the point cloud acquired from scanning experiments further validated the MSL system’s performance. This finding preliminarily validates the feasibility of multispectral LiDAR for classifying submerged spherical targets.

In order to deal with the target recognition in the complex underwater environment, we carried out experimental research. This includes filtering noise in the feature extraction stage of underwater images rich in noise, or with complex backgrounds, and improving the accuracy of target classification in the recognition process. This paper discusses our contribution to improving the accuracy of underwater target classification. This paper proposes an underwater target classification algorithm based on the improved flow direction algorithm (FDA) and search agent strategy, which can simultaneously optimize the weight parameters, bias parameters, and super parameters of the extreme learning machine (ELM). As a new underwater target classifier, it replaces the full connection layer in the traditional classification network to build a classification network. In the first stage of the network, the DenseNet201 network pre-trained by ImageNet is used to extract features and reduce dimensions of underwater images. In the second stage, the optimized ELM classifier is trained and predicted. In order to weaken the uncertainty caused by the random input weight and offset of the introduced ELM, the fuzzy logic, chaos initialization, and multi population strategy-based flow direction algorithm (FCMFDA) is used to adjust the input weight and offset of the ELM and optimize the super parameters with the search agent strategy at the same time. We tested and verified the FCMFDA-ELM classifier on Fish4Knowledge and underwater robot professional competition 2018 (URPC 2018) datasets, and achieved 99.4% and 97.5% accuracy, respectively. The experimental analysis shows that the FCMFDA-ELM underwater image classifier proposed in this paper has a greater improvement in classification accuracy, stronger stability, and faster convergence. Finally, it can be embedded in the recognition process of underwater targets to improve the recognition performance and efficiency.

The ocean plays an important role in maintaining the equilibrium of Earth’s ecology and providing humans access to a wealth of resources. To obtain a high-precision underwater image classification model, we propose a classification model that combines an EfficientnetB0 neural network and a two-hidden-layer random vector functional link network (EfficientnetB0-TRVFL). The features of underwater images were extracted using the EfficientnetB0 neural network pretrained via ImageNet, and a new fully connected layer was trained on the underwater image dataset using the transfer learning method. Transfer learning ensures the initial performance of the network and helps in the development of a high-precision classification model. Subsequently, a TRVFL was proposed to improve the classification property of the model. Net construction of the two hidden layers exhibited a high accuracy when the same hidden layer nodes were used. The parameters of the second hidden layer were obtained using a novel calculation method, which reduced the outcome error to improve the performance instability caused by the random generation of parameters of RVFL. Finally, the TRVFL classifier was used to classify features and obtain classification results. The proposed EfficientnetB0-TRVFL classification model achieved 87.28%, 74.06%, and 99.59% accuracy on the MLC2008, MLC2009, and Fish-gres datasets, respectively. The best convolutional neural networks and existing methods were stacked up through box plots and Kolmogorov-Smirnov tests, respectively. The increases imply improved systematization properties in underwater image classification tasks. The image classification model offers important performance advantages and better stability compared with existing methods.

The classification of shallow sea sediments based on airborne LiDAR bathymetry represents a significant advancement in marine science and engineering. Airborne LiDAR is a highly valuable tool for the classification of seabed sediments, offering high accuracy and mobility. However, accurately classifying shallow marine sediments remains a challenging endeavor due to the difficulties associated with differentiation and the inherent limitations in accuracy. To achieve the accurate classification of underwater sediments, a feature selection method for underwater sediment classification is proposed in this paper and tested in a laboratory environment. The method inputs the original feature set into a classification algorithm that combines Sequential Forward Selection with Random Forests. The study demonstrates that the model achieves an overall classification accuracy of 94.1% and a Kappa coefficient of 91.11%, thereby enabling the accurate and efficient classification of underwater sediment. This approach can be employed as a supplementary technique for the precise classification of shallow marine sediments, offering valuable assistance in the examination of marine ecosystems.