REFNet: reparameterized feature enhancement and fusion network for underwater blur target recognition

Objective: We propose a deep-learning-based underwater target detection system that can effectively solve the problem of underwater optical image target detection and recognition. Methods: In this paper, based on the depth of the underwater optical image target detection and recognition and using a learning model, we put forward corresponding solutions using the concept of style migration solutions, such as training samples. A lack of variability and poor generalization of practical applications presents a challenge for underwater object identification. The UW_YOLOv3 lightweight model was proposed to solve the problems of calculating energy consumption and storage resource limitations in underwater application scenarios. The detection and recognition module, based on deep learning, can deal with the degradation process of underwater imaging by embedding an image enhancement module into the detection and recognition module for the joint tuning and transferring of knowledge. Results: The detection accuracy of the UW_YOLOv3 model designed in this paper outperformed the lightweight algorithm YOLOV3-TINY by 7.9% at the same image scale input. Compared with other large algorithms, the detection accuracy was lower, but the detection speed was much higher. Compared with the SSD algorithm, the detection accuracy was only 4.7 lower; the speed was 40.9 FPS higher; and the rate was nearly 16 times higher than Faster R-CNN. When the input scale was 224, although part of the accuracy was lost, the detection speed doubled, reaching 156.9 FPS. Conclusion: Based on our framework, the problem of underwater optical image target detection and recognition can be effectively solved. Relevant studies have not only enriched the theory of target detection and glory, but have also provided optical glasses with a clear vision for appropriate underwater application systems.

In view of the problem of difficult feature extraction and target missed detection, propose an improved transformer based underwater target detection algorithm. The proposed algorithm improves the linear embedding module of the Transformer to supplement the linear embedding information of small objects and builds the local relationship model of each layer feeling field, in order to achieve end-to-end accurate detection. An experiment on URPU underwater target detection data set is conducted. Compared with Swim Transformer, the proposed algorithm detection accuracy is 77.4%, 1. 5%, detection speed is 7 frame/s, 35.6%, the number is 30.4 MB, and compression is 84.1%. Experimental results have shown that the improved algorithm improves the performance of underwater small target detection

Underwater target detection is an indispensable part of marine environmental engineering and a fast and accurate method of detecting underwater targets is essential. Although many target detection algorithms have achieved great accuracy in daily scenes, there are issues of low-quality images due to the complex underwater environment, which makes applying these deep learning algorithms directly to process underwater target detection tasks difficult. In this paper, we presented an algorithm for underwater target detection based on improved You Only Look Once (YOLO) v4 in response to the underwater environment. First, we developed a new convolution module and network structure. Second, a new intersection over union loss was defined to substitute the original loss function. Finally, we integrated some other useful strategies to achieve more improvement, such as adding one more prediction head to detect targets of varying sizes, integrating the channel attention into the network, utilizing K-means++ to cluster anchor box, and utilizing different activation functions. The experimental results indicate that, in comparison with YOLOv4, our proposed algorithm improved the average accuracy of the underwater dataset detection by 10.9%, achieving 91.1%, with a detection speed of 58.1 frames per second. Therefore, compared to other mainstream target detection algorithms, it is superior and feasible for applications in intricate underwater environments.

The underwater environment is more complex than that on land, resulting in severe static and dynamic blurring in underwater images, reducing the recognition accuracy of underwater targets and failing to meet the needs of underwater environment detection. Firstly, for the static blurring problem, we propose an adaptive color compensation algorithm and an improved MSR algorithm. Secondly, for the problem of dynamic blur, we adopt the Restormer network to eliminate the dynamic blur caused by the combined effects of camera shake, camera out-of-focus and relative motion displacement, etc. then, through qualitative analysis, quantitative analysis and underwater target detection on the enhanced dataset, the feasibility of our underwater enhancement method is verified. Finally, we propose a target recognition network suitable for the complex underwater environment. The local and global information is fused through the CCBC module and the ECLOU loss function to improve the positioning accuracy. The FasterNet module is introduced to reduce redundant computations and parameter counting. The experimental results show that the CFEC-YOLOv7 model and the underwater image enhancement method proposed by us exhibit excellent performance, can better adapt to the underwater target recognition task, and have a good application prospect.

The performance of underwater target detection algorithms is affected by poor imaging quality in underwater environments. Due to the arithmetic power limitation of underwater devices, existing deep learning networks are unable to provide efficient detection processes with high detection accuracy. Lightweight CNN models have been actively applied for underwater environment detection, yet their lite feature fusion networks cannot provide effective fusion effects and reduce the detection accuracy. In this paper, a lightweight algorithm based on multi-scale feature fusion was proposed, with the model parameters greatly reduced, improving the target detection accuracy. The forward propagation memory overhead is reduced by using multi-scale shared convolutional kernels and pooling operations to co-construct the query matrix in the Tansformer encoding stage. Then, the feature fusion path is optimized in order to enhance the connection of multi-scale features. A multiscale feature adaptive fusion strategy is used to enhance the detection performance and reduce the dependence on the complex feature extraction network. The feature extraction network is also reparameterized to simplify the operation. Using the UPRC offshore dataset for validation, the study results have demonstrated that the statistical mAP metrics validate the detection accuracy. Compared with SSD, RetinaNet and YOLOv5-s improved by 13%, 8.6%, and 0.8%, while the number of parameters decreased by 76.09%, 89.74%, and 87.67%. In addition, compared with the YOLOv5-lite model algorithm with the same parameter volume, the mAP is improved by 3.8%, which verifies the accuracy and efficiency of the algorithm in this paper.

The detection and recognition of underwater preset structure is an important topic in the detection of potential safety hazards in underwater engineering. At present, the detection of underwater preset structure is mainly based on multi-beam sounding technology. In this paper, the submarine pipeline is taken as the research object, and a new detection method based on 3D real-time imaging sonar technology is proposed, which realizes the underwater environment acoustic high-precision real-time target detection and fine target analysis. First of all, the underwater target detection data is obtained by using 3D imaging sonar, high-precision integrated navigation, tide gauge, CTD and other equipment. Secondly, the attitude correction, sound velocity correction and depth compensation of 3D imaging data is carried out, and the bilateral filtering combined with threshold filtering is used to filter the data. Finally, combined with advanced point cloud data processing algorithms and edge data extraction based on eigenvalues, accurate identification, positioning and status analysis of submarine pipelines are achieved. The results of practical engineering detection show that this method has the advantages of high resolution, real-time imaging and excellent imaging quality. The detection image is stereoscopic and clear, which can provide more detailed description of underwater target contour. The data processing and analysis method adopted in this paper can track and identify the submarine pipeline quickly and accurately, and detect the areas with potential safety hazards such as exposed or suspended pipelines, which provides scientific basis for the timely management of pipelines. This paper provides a new detection method for submarine pipeline detection, which has applicability and popularizing value.

Underwater robots are an important tool for exploiting marine resources, while underwater target detection is a key technology for marine fishing by underwater robots. For the problem that the traditional algorithm is difficult to detect small targets and multiple scales in the underwater environment, this paper proposes an underwater biological detection algorithm based on the multi-scale feature fusion and attention mechanism. First, a YOLOv5 network is constructed, and a bi-directionally weighted BiFPN feature pyramid structure is fed into the neck network so that multiscale target features can be efficiently merged. Second, a channel and spatial convolutional attention mechanism is introduced between the neck network and the prediction network in order to increase the saliency of the underwater target that needs to be detected, which effectively reduces the missed and erroneous detection of underwater targets. The experimental results show that the proposed algorithm can accurately detect multi-scale underwater targets in complex underwater environments with an average accuracy of 80.18%, which is 2.87% better than YOLOv5, and an average recall of 85.06%, which is 4.81% better than YOLOv5.

The underwater environment of the sea bottom is complex, and it is easy to be affected by light, underwater impurities and shooting angle. Because of the dense and mutual occlusion of underwater targets, the collected images are not clear, fuzzy and easy to be resolved, which affects the accuracy of target detection and recognition. Aiming at the above problems, this paper proposes an improved YOLOV5 method for underwater seafood target detection. Firstly, the C3 module of high-level features is fused with swin transformer to create the C3STR module, and the improved C3STR module can enlarge the receptive field of the network, obtain the global information better and increase the ability of feature extraction, the DC3 module is constructed by adding convolution to the low-level feature C3 module, which improves the model detection ability with a little more computation. Finally, the network feature fusion module is improved, which integrates the deep-level information with the shallow-level information, the combination of semantic information and position information can improve the precision of model recognition and enhance the ability of target detection. The experimental results show that the detection accuracy of the improved YOLOv5 algorithm is increased to 88.3%, which is 2% higher than that of the original YOLOv5 algorithm.

Neuronal morphology can be represented using various feature representations, such as hand-crafted morphometrics and deep features. These features are complementary to each other, contributing to improving performance. However, existing classification methods only utilize a single feature representation or simply concatenate different features without fully considering their complementarity. Therefore, their performance is limited and can be further improved. In this paper, we propose a multi-level feature fusion network that fully utilizes diverse feature representations and their complementarity to effectively describe neuronal morphology and improve performance. Specifically, we devise a Multi-Level Fusion Module (MLFM) and incorporate it into each feature extraction block. It can facilitate the interaction between different features and achieve effective feature fusion at multiple levels. The MLFM comprises a channel attention-based Feature Enhancement Module (FEM) and a cross-attention-based Feature Interaction Module (FIM). The FEM is used to enhance robust morphological feature presentations, while the FIM mines and propagates complementary information across different feature presentations. In this way, our feature fusion network ultimately yields a more distinctive neuronal morphology descriptor that can effectively characterize neurons than any singular morphological representation. Experimental results show that our method effectively depicts neuronal morphology and correctly classifies 10-type neurons on the NeuronMorpho-10 dataset with an accuracy of 95.18%, outperforming other approaches. Moreover, our method performs well on the NeuronMorpho-12 and NeuronMorpho-17 datasets and possesses good generalization.

Underwater biological detection is of great significance to marine protection. However, the traditional target detection techniques have some challenges, such as insufficient feature extraction for small targets and low feature utilization rate. To address these challenges, an underwater biological detection model IG-YOLOv5 based on the idea of feature reuse is proposed. An Improved-Ghost module with feature reuse is designed. The module adds batch normalization operations to the identity mapping branch using the Add operation with feature fusion and the Sigmoid Linear Unit activation function with smoother zeros. The proposed model uses the Improved-Ghost module to reconstruct the CSPDarknet structure of YOLOv5, so as to realize the lightweight and accuracy improvement of the model. In addition, in order to solve the problem of target size and shape change in underwater environment, the optimized loss function is Wise-IoU v3, which is used to evaluate the accuracy and robustness of detection results. The results show that the IG-YOLOv5 model performs well in the 2021URPPC data set, with 0.5 mAP reaching 74.2, 4.3% higher than that of YOLOv5 model, and 2.7 less floating-point operations. In a word, IG-YOLOv5 model has high accuracy and robustness in underwater target detection, and Wise-IoU index can evaluate the quality of target detection results more accurately, which is suitable for underwater robots, underwater monitoring, and other fields and has a practical application value.

The classification of high-resolution (HR) synthetic aperture radar (SAR) images is of great importance for SAR scene interpretation and application. However, the presence of intricate spatial structural patterns and complex statistical nature makes SAR image classification a challenging task, especially in the case of limited labeled SAR data. This paper proposes a novel HR SAR image classification method, using a multi-scale deep feature fusion network and covariance pooling manifold network (MFFN-CPMN). MFFN-CPMN combines the advantages of local spatial features and global statistical properties and considers the multi-feature information fusion of SAR images in representation learning. First, we propose a Gabor-filtering-based multi-scale feature fusion network (MFFN) to capture the spatial pattern and get the discriminative features of SAR images. The MFFN belongs to a deep convolutional neural network (CNN). To make full use of a large amount of unlabeled data, the weights of each layer of MFFN are optimized by unsupervised denoising dual-sparse encoder. Moreover, the feature fusion strategy in MFFN can effectively exploit the complementary information between different levels and different scales. Second, we utilize a covariance pooling manifold network to extract further the global second-order statistics of SAR images over the fusional feature maps. Finally, the obtained covariance descriptor is more distinct for various land covers. Experimental results on four HR SAR images demonstrate the effectiveness of the proposed method and achieve promising results over other related algorithms.

Magnetic detection is an effective way for underwater buried target detection especially man-made objects which will contain some amount of permeable materials. Magnetic sensing can augment acoustic means of detection, effectively reduce false alarms. The advances in Autonomous Underwater Vehicle (AUV) technology enabled the integration of magnetic gradiometer onboard for underwater buried object detection. In this paper, the design considerations and configurations of a prototype magnetic gradiometer integrated to an AUV and the magnetic data fusion with AUV navigation data will be presented. The magnetic gradiometer formed by fluxgate magnetometers has several advantages such as small size and low power with relatively high sensitivity. Magnetic noise compensation is also critical as the magnetometer has to operate close to the AUV magnetic noise sources. The paper will describe the results of the platform magnetic noise compensation and sea trial measurements.

For the detection of shallow underwater targets by the hyperspectrum, the information of the target is carried in the water-leaving radiation. However, the differences between the water-leaving radiation of the underwater target and the ocean background radiation are not significant, and the spectrum of the water-leaving radiation is sensitive to the changes of the marine environment. Therefore, taking the spectrum as the feature is not stable for hyperspectral detection of the underwater target. In this paper, we propose a parameter feature extraction method for hyperspectral detection of the shallow underwater target. The presence of an underwater target would change the optical properties of seawater in a local place. Based on that, a quasi-analytical model of the water-leaving radiation is adopted to inverse the properties of the marine environment from the radiated spectrum. Then the inversed water depth and bottom reference reflectance are taken as the features. Simulation results show that the samples which belong to the water-leaving radiation of underwater target have higher inversed depth and reference reflectance than that belong to the ocean background radiation. Furthermore, these two kinds of samples can be separated significantly and stably in a 2-dimension feature space.

The most popular and crucial piece of equipment for underwater target detection right now is side-scan sonar, which primarily plays a crucial role in underwater operations. Unfortunately, in general, side-scan sonar images have low resolution and suffer from significant noise interference, making target recognition with these images highly challenging. In this paper, we employ deep learning techniques to recognize underwater targets using side scan sonar imagery. This paper begins by outlining the current state of underwater target detection research as well as briefly describing its importance and historical context. The imaging process of side-scan sonar pictures is studied in greater depth, and it is determined that noise on sonar images is caused principally by the combined effect of external and internal noise. This causes extra noise in sonar pictures, lowering image quality and affecting target identification performance. Proper sonar image pre-processing findings can provide more consistent target detection help. After then, the YOLOv5 target detection algorithm is investigated and improved. By performing several sets of comparison trials, the updated technique improves the accuracy and speed of side-scan sonar image target recognition and finally overcomes the issues of missed detection, false detection, and low accuracy of overlapping tiny object recognition.

We consider underwater target detection by using a novel active, self-contained and rapidly deployable underwater inflatable co-prime sonar array (UICSA). In particular, we measure the received signal strength (RSS) and angle-of-arrival (AoA) of acoustic signals reflected by an underwater target. Measurements from different positions of the transmitter with respect to the UICSA are organized in a three mode real-valued tensor. Then, the conformity of each entry with respect to all other data points in the tensor is calculated based on recursively refined calculations of $L_{1}$ -norm tensor subspaces. Conformity values are then used for the detection of an underwater target. We evaluate the data conformity of RSS and AoA recordings acquired from the testbed deployment of a seven-element UICSA prototype and three underwater acoustic transmitters at Florida Atlantic University. We show for the first time that conformity evaluation over multi-modal data measurements can accurately detect the presence of an underwater target.