The coastal zone is connected to the hinterland of the basin and the wide sea, which is not only affected by complex natural factors, but also by human activities. The study of sediment classification in this area will help to further explain topographic evolution and dynamic mechanism. Aiming at the complex mixed sediment environment in the ancient valley area of shallow seas, an advanced deep convolutional neural network classification model based on transfer learning—selective kernel hybrid dilated ResNet-50 (SKHD-ResNet-50) is thus proposed and applied to the coastal area in the Firth of Forth, Scotland, UK. The model overcomes problems of low accuracy of sediment classification with small samples, while effectively improving the efficiency of mixed sediment classification. Five transferred convolutional neural network models including CNN, AlexNet, VGG-19, GoogleNet, and SKHD-ResNet-50, were used to conduct seabed sediment classification experiments on multibeam backscatter mosaics in the study area. An ablation study about the effect of selective kernel convolution (SKC) and hybrid dilated convolution (HDC) was also conducted to further prove the usability of the proposed method. Finally, the optimized convolutional neural network was used to build a multi-layer deep learning model for sediment classification, combining the transfer learning method with the neural network to obtain the optimal network weights through deep global optimization, thus accurately mining statistical characteristics and distribution rules of backscatter intensity data. The experimental results show that the advanced model obtains better classification accuracy than that of state-of-art methods, with overall accuracy and Kappa coefficient of 96.87% and 0.9482, respectively. Findings highlight the proposed sediment classification model, which can effectively distinguish mixed sediments, and greatly improve the accuracy and efficiency of sediment classification.

Precision bottom tracking is a core step in the data processing of side scan sonar (SSS), which is of critical value for the quality of the final SSS production. Currently, the automatic precision SSS bottom tracking remains challenging due to the complex noise caused by the measuring environment; especially, the existing methods did not fully exploit the temporal correlations or depended on the hand-crafted setting. Therefore, we proposed a novel temporal fusion-based 1-D sequence semantic segmentation model, TFSSM–1–D, to fuse the temporal correlation features and perform automatic precision SSS bottom tracking. The TFSSM–1–D uses the deep learning (DL) encoder–decoder model for mapping inputs to 1-D semantic label outputs, with the aid of preprocess and temporal fusion modules, to improve the accuracy and robustness. Among them, preprocess module is used to introduce the prior knowledge of bilateral symmetry, which alleviates the defect of long-distance features correlation caused by the inductive bias of convolution neural network (CNN). The temporal fusion consists of the point-wise temporal fusion module (PTFM), and the bi-directional attention propagation module (BAPM) guides the model to explicitly fuse the temporal variation features on different scales. The experimental results demonstrate the effectiveness of TFSSM–1–D, and its mean port offset error (MPOE) reaches 2.7058 on the testing set without downsampling, which is 40% lower than the previous DL based model with single ping input, and other evaluation metrics have also significantly improved. The inference on unseen data shows that TFSSM–1–D can achieve precision bottom tracking with good noise immunity.

The bottom tracking is to confirm the first bottom return signal in each ping to divide the data into the seabed area and the water column area, which is an essential part of side-scan sonar (SSS) data processing. More than that, SSS is usually affected by water flow or suspended solids, resulting in serious noises, which make the bottom tracking more difficult and complex. Some existing methods follow ideas such as abnormal point detection or filtering, resulting in unsatisfactory bottom tracking effects. Inspired by the semantic segmentation mission in deep learning, we innovatively transform the bottom tracking into a binary-class semantic segmentation mission on 1-D SSS data. Following this idea, we propose a semantic segmentation-based bottom tracking method and modify two image semantic segmentation models, SegNet and U-Net, so that they can achieve bottom tracking on 1-D SSS data. Benefiting from the large receptive field brought by convolution and down-sampling, our method has made significant progress in accuracy and antinoise performance compared with the existing method. According to the experimental result, our method has obvious improvement compared with the existing method on the testing set; not only that, our method also performs better on the unused data.

The ability to accurately map the seabed sediments plays an important role in seabed habitat development and stakeholder decision-making. In conventional seabed sediment classification methods, maps of seabed sediment are provided in categorical form (sediment classes). Therefore, the prediction of the sediment compositions in multibeam observational units has become a difficult issue in using conventional methods. To tackle this challenge, a new strategy is developed to realize the subpixel decomposition of seabed sediments. A key attribute of the proposed sediment decomposition model is that it utilizes spatial–spectral information provided by multibeam backscatter angular responses (ARs). First, an AR feature extraction method utilizing a bidirectional sliding window is proposed and a$K$-means clustering algorithm is used for segmentation. Second, a deep-sea sediment decomposition model based on the fuzzy method is constructed by selecting experimental samples that are distributed within a single clustering region. This model inverts the abundance of each sediment composition in the form of membership degrees. Finally, deep-sea multibeam survey data collected from the central Philippine Sea are used for verification. The overall mean square error and coefficient of determination reach 0.043 and 0.856, respectively. The experimental results show that the new method can accurately decompose deep-sea sediment compositions, thus providing a new technique for deep-sea acoustic sediment remote sensing and quantitative analysis.

For mapping the near-shore seafloor bathymetry, retrieving depth information using multispectral satellite image is highly cost-effective. To effectively detect and characterize the bathymetry variation, accurate and reliable information about the uncertainty of the derived depths is critical. In estimating the uncertainty of the resulted satellite-derived bathymetry (SDB), the conventional homoscedasticity assumption states that the error variance of the observations is constant across different depth ranges. However, this assumption is violated due to the influence of various environmental factors inherently correlated with depth. In this article, we develop a data-driven approach to extract the depth-dependent pattern of observation error. The residual information of the regression is analyzed to model the influence of the depth on the uncertainty of retrieval bathymetry, while nonlinearity and outliers are also considered. This results in a more realistic estimate of SDB accuracy. Our experimental results reveal that the observation uncertainty is significantly correlated with the depth in the bathymetry retrieval process. It is also shown that the presented algorithm effectively captures the depth-dependent pattern of the observation uncertainty, and further provides a more realistic characterization of the uncertainty information of SDB.

An algorithm is described and tested to provide accurate and robust deep-sea seafloor classification based on the backscatter data derived from a multibeam bathymetry system. This article focuses on significant heterogeneity in the deep-sea backscatter strength (BS) data. The angular response curve information is decomposed into different units, and BS data are grouped on the basis of the incidence angle to address the heterogeneity in the across-ship direction. Subsequently, a sliding window is applied on BS data in each group, and a robust estimation method is used to address the potential heterogeneity in the window during feature extraction. Thereafter, the extracted features are learned by fuzzy c-means (FCM) to obtain a clustering solution. In the learning process, the features of each group are learned by an independent FCM. The modified FCM algorithm is used to cluster each group of data to handle unbalanced backscatter data sets. With this procedure, heterogeneity in BS data can be accounted for, which is universal in deep-sea survey application. Finally, the results of the different groups are merged to obtain a global label set for the survey region. The method was tested on the multibeam data collected from an offshore region around the Kyushu-Palau Ridge. Monte Carlo simulation was performed to evaluate the performance of the robust method. Computational results demonstrate that the improved algorithm can address the heterogeneity in BS data efficiently and provide an accurate classification solution in the deep-sea survey environment.

Mainly due to light absorption and scattering, underwater images usually suffer from low contrast, blur and color cast. An underwater image restoration method based on color correction and Red Channel prior is proposed in this paper. Firstly, the Retinex algorithm is adopted to eliminate the uneven illumination. Secondly, the image is compensated based on the underwater attenuation imaging model. The color of the light source is estimated and eliminated to achieve color correction. Finally, the Red Channel prior method is used for image deblurring. Compared with other methods, experimental results demonstrate that the proposed method can effectively eliminate the unevenness of illumination and color distortion, and improve the clarity and contrast, while removing the blur of the underwater color image.

Designing maximally selective ligands that act on individual drug targets with high binding affinity has been the central dogma of drug discovery and development for the past two decades. However, many low-affinity drugs that aim for several targets at the same time are found more effective than the high-affinity binders when faced with complex disease conditions, such as cancers, Alzheimer's disease and cardiovascular diseases. The aim of this study was to appreciate the importance and reveal the features of weak-binding drugs and propose an integrated strategy for discovering them. Weak-binding drugs can be characterized by their high dissociation rates and transient interactions with their targets. In addition, network topologies and dynamics parameters involved in the targets of weak-binding drugs also influence the effects of the drugs. Here, we first performed a dynamics analysis for 33 elementary subgraphs to determine the desirable topology and dynamics parameters among targets. Then, by applying the elementary subgraphs to the mitogen-activated protein kinase (MAPK) pathway, several optimal target combinations were obtained. Combining drug-target interaction prediction with molecular dynamics simulation, we got two potential weak-binding drug candidates, luteolin and tanshinone IIA, acting on these targets. Further, the binding affinity of these two compounds to their targets and the anti-inflammatory effects of them were validated through in vitro experiments. In conclusion, weak-binding drugs have real opportunities for maximum efficiency and may show reduced adverse reactions, which can offer a bright and promising future for new drug discovery.

MOTIVATION: Drug combinations are a promising strategy for combating complex diseases by improving the efficacy and reducing corresponding side effects. Currently, a widely studied problem in pharmacology is to predict effective drug combinations, either through empirically screening in clinic or pure experimental trials. However, the large-scale prediction of drug combination by a systems method is rarely considered. RESULTS: We report a systems pharmacology framework to predict drug combinations (PreDCs) on a computational model, termed probability ensemble approach (PEA), for analysis of both the efficacy and adverse effects of drug combinations. First, a Bayesian network integrating with a similarity algorithm is developed to model the combinations from drug molecular and pharmacological phenotypes, and the predictions are then assessed with both clinical efficacy and adverse effects. It is illustrated that PEA can predict the combination efficacy of drugs spanning different therapeutic classes with high specificity and sensitivity (AUC = 0.90), which was further validated by independent data or new experimental assays. PEA also evaluates the adverse effects (AUC = 0.95) quantitatively and detects the therapeutic indications for drug combinations. Finally, the PreDC database includes 1571 known and 3269 predicted optimal combinations as well as their potential side effects and therapeutic indications. AVAILABILITY AND IMPLEMENTATION: The PreDC database is available at http://sm.nwsuaf.edu.cn/lsp/predc.php.