Complex Underwater Environment Research Articles

Research Progress of Path Planning Methods for Autonomous Underwater Vehicle

Path planning is a key technology for autonomous underwater vehicle (AUV) navigation. With the emphasis and research on AUV, AUV path planning technology is continuously developing. Path planning techniques generally include environment modelling methods and path planning algorithms. Based on a brief description of the environment modelling methods, this paper focuses on the path planning algorithms commonly used by AUV. According to the basic principles of the algorithm, the AUV path planning algorithms are divided into four categories: artificial potential field methods, geometric model search methods, random sampling methods, and intelligent bionic methods. In this review, we summarize in detail the development and application of various path planning algorithms in recent years. Meanwhile, we analyse the advantages and disadvantages of various algorithms and their improvement methods. Obstacles, ocean currents, and undersea terrain have an impact on AUV path planning. Therefore, how to deal with the complex underwater environment adds some limits to AUV path planning algorithms. In addition to the external environment, path planning algorithms also need to consider AUV’s physical constraints, such as energy constraints and motion constraints. Then, we analyse the motion constraints in AUV path planning. Finally, we discuss the development direction of AUV path planning algorithm. Time-varying ocean currents, special obstacles, multiobjective constraints, and practicability will be the problems that AUV path planning algorithms need to solve.

Study on Time Reversal Maximum Ratio Combining in Underwater Acoustic Communications

Time reversal (TR) can achieve temporal and spatial focusing by exploiting spatial diversity in complex underwater environments with significant multipath. This property makes TR useful for underwater acoustic (UWA) communications. Conventional TR is realized by performing equal gain combining (EGC) on the single element TR output signals of each element of the vertical receive array (VRA). However, in the actual environment, the signal-to-noise ratio (SNR) and the received noise power of each element are different, which leads to the reduction of the focusing gain. This paper proposes a time reversal maximum ratio combining (TR-MRC) method to process the received signals of the VRA, so that a higher output SNR can be obtained. The theoretical derivation of the TR-MRC weight coefficients indicates that the weight coefficients are only related to the input noise power of each element, and are not affected by the multipath structure. The correctness of the derivation is demonstrated with the experimental data of the long-range UWA communications conducted in the South China Sea. In addition, the experimental results illustrate that compared to the conventional TR, TR-MRC can provide better performance in terms of output SNR and bit error rate (BER) in UWA communications.

Composited FishNet: Fish Detection and Species Recognition From Low-Quality Underwater Videos

The automatic detection and identification of fish from underwater videos is of great significance for fishery resource assessment and ecological environment monitoring. However, due to the poor quality of underwater images and unconstrained fish movement, traditional hand-designed feature extraction methods or convolutional neural network (CNN)-based object detection algorithms cannot meet the detection requirements in real underwater scenes. Therefore, to realize fish recognition and localization in a complex underwater environment, this paper proposes a novel composite fish detection framework based on a composite backbone and an enhanced path aggregation network called Composited FishNet. By improving the residual network (ResNet), a new composite backbone network (CBresnet) is designed to learn the scene change information (source domain style), which is caused by the differences in the image brightness, fish orientation, seabed structure, aquatic plant movement, fish species shape and texture differences. Thus, the interference of underwater environmental information on the object characteristics is reduced, and the output of the main network to the object information is strengthened. In addition, to better integrate the high and low feature information output from CBresnet, the enhanced path aggregation network (EPANet) is also designed to solve the insufficient utilization of semantic information caused by linear upsampling. The experimental results show that the average precision (AP)0.5:0.95, AP50 and average recall (AR)max=10 of the proposed Composited FishNet are 75.2%, 92.8% and 81.1%, respectively. The composite backbone network enhances the characteristic information output of the detected object and improves the utilization of characteristic information. This method can be used for fish detection and identification in complex underwater environments such as oceans and aquaculture.

An Underwater Integrated Navigation Algorithm to Deal With DVL Malfunctions Based on Deep Learning

In underwater navigation systems, Global Navigation Satellite System (GNSS) information cannot be used for navigation. The mainstream method of autonomous underwater vehicles (AUV) underwater navigation system is Doppler Velocity Log (DVL) aided strapdown inertial navigation system (SINS). However, because the DVL is an instrument based on Doppler frequency shift to measure velocity, it is easily affected by the external environment. In a complex underwater environment, DVL output is easily polluted by outliers or even interrupted. In this paper, A new integrated navigation algorithm based on deep learning model is proposed to deal with DVL malfunctions. First, use RKF based on Mahalanobis distance algorithm to eliminate outliers, and then train the Nonlinear AutoRegressive with eXogenous input (NARX) model when DVL is available. When DVL is interrupted, use the NARX model to predict the output of DVL and continue integrated navigation. The proposed NARX-RKF scheme’s effectiveness verification was performed on the data set collected by the SINS/DVL ship-mounted experimental system. For comparison, different methods are also compared in the experiment. Experimental results show that NARX-RKF can effectively predict the output of DVL and is significantly better than other methods.

An Underwater Single Target Tracking Method Using SiamRPN++ Based on Inverted Residual Bottleneck Block

Traditional tracking algorithms reply on manually extracted features while deep learning algorithms can automatically extract features, which is of great importance for single target tracking in the complex underwater environment that is characterized by poor visibility, low contrasts, and occlusion. The convolutional neural network SiamRPN++ is an advanced deep learning algorithm whose backbone network structure is. However, this algorithm is difficult to be implemented in an underwater platform due to its large amount of computation and high requirement for the computing power of the hardware. To solve this problem, a new backbone network structure NewNet-62 characterized by inverted residual bottleneck block and the SiamRPN++ algorithm are proposed. The depth-wise separable convolution in NewNet-62 simplifies the computation so that the SiamRPN++ algorithm can realize fast and accurate tracking of underwater single targets. The video set in VOT2016 was used to test the algorithm. The results show that the SiamRPN++ algorithm using NewNet-62 as the main network has the great performance. Compared with the algorithms using ResNet-50, the accuracy of the SiamRPN++ algorithm increased to 0.524 and its EAO increased to 0.303. And its tracking speed increased to 73.74. The network complexity reduced to 3.013 billion and the network parameters reduced to 12.538 million, which significantly reduces network complexity and number of network parameters.

Underwater Distortion Target Recognition Network (UDTRNet) via Enhanced Image Features.

It is difficult for the autonomous underwater vehicle (AUV) to recognize targets similar to the environment in lacking data labels. Moreover, the complex underwater environment and the refraction of light cause the AUV to be unable to extract the complete significant features of the target. In response to the above problems, this paper proposes an underwater distortion target recognition network (UDTRNet) that can enhance image features. Firstly, this paper extracts the significant features of the image by minimizing the info noise contrastive estimation (InfoNCE) loss. Secondly, this paper constructs the dynamic correlation matrix to capture the spatial semantic relationship of the target and uses the matrix to extract spatial semantic features. Finally, this paper fuses the significant features and spatial semantic features of the target and trains the target recognition model through cross-entropy loss. The experimental results show that the mean average precision (mAP) of the algorithm in this paper increases by 1.52% in recognizing underwater blurred images.

Feature Cascade Underwater Object Detection Based on Stereo Segmentation

Kong, W.; Yang, M., and Huang, Q., 2020. Feature cascade underwater object detection based on stereo segmentation. In: Liu, X. and Zhao, L. (eds.), Today's Modern Coastal Society: Technical and Sociological Aspects of Coastal Research. Journal of Coastal Research, Special Issue No. 111, pp. 140–144. Coconut Creek (Florida), ISSN 0749-0208.Underwater object detection is a hot topic in deep learning in recent years. It is difficult to quickly complete the task of object detection due to the complex underwater environment. This article presents a novel neural network architecture, namely, Feature Cascade Underwater Object Detection Network. First, the estimated weights of different scales are obtained through the key region detection module. Then in the second module, the complete segmentation image is obtained through the multiple scale segmentation method. Third, cascade multilevel weights are estimated, along with stereo segmentation, for layer by layer integration. Finally, the predicted underwater object detection results. We test our network on the URPC2017 and LABELED-FISHES-IN-THE-WILD dataset. The average accuracy of our algorithm is higher than the other algorithms in single-category and multicategory underwater object detection tasks. Especially in the fish object detection task, our network obtains an average accuracy value that reaches 93.4%.

Adaptive Histogram Fusion-based Color Restoration and Enhancement for Underwater Images

Underwater image restoration is crucial for marine exploration, underwater archaeology and seafloor geological investigation. However, due to the complex underwater environment, the image obtained underwater often has problems with colour deviation and lack of details. To improve the quality and the visual effect of underwater images, we propose an adaptive histogram fusion-based colour restoration and enhancement method for underwater images. First, the proposed method restores the colour by pixel value adjusting in a channel and b channel according to colour distortion. Then the histogram is used in the H channel of the underwater source image to improve the colour hue. To make the enhanced result more in line with visual observation, the edge reserving-based method is used to enhance the details of colour restoration results. Experimental results demonstrate that our results are effective in subjective and objective evaluations, and superior to several existing methods.

Improved exponential weighted moving average based measurement noise estimation for strapdown inertial navigation system/doppler velocity log integrated system

AbstractThe strapdown inertial navigation system (SINS) with integrated Doppler velocity log (DVL) is widely utilised in underwater navigation. In the complex underwater environment, however, the DVL information may be corrupted, and as a result the accuracy of the Kalman filter in the SINS/DVL integrated system degrades. To solve this, an adaptive Kalman filter (AKF) with measurement noise estimator to provide noise statistical characteristics is generally applied. However, existing methods like moving windows (MW) and exponential weighted moving average (EWMA) cannot adapt to a dynamic environment, which results in unsatisfactory noise estimation performance. Moreover, the forgetting factor has to be determined empirically. Therefore, this paper proposes an improved EWMA (IEWMA) method with adaptive forgetting factor for measurement noise estimation. First, the model for a SINS/DVL integrated system is established, then the MW and EWMA based measurement noise estimators are illustrated. Subsequently, the proposed IEWMA method which is adaptive to the various environments without experience is introduced. Finally, simulation and vehicle tests are conducted to evaluate the effectiveness of the proposed method. Results show that the proposed method outperforms the MW and EWMA methods in terms of measurement noise estimation and navigation accuracy.

Underwater broadband sound insulation with chiral spiral structures

Low-frequency noise has been a long-standing issue in the context of a complex underwater environment. While theory related to low-frequency sound insulation in air has been developed, systematic research on underwater sound insulation has not yet been established. In our work, we used a chiral variable pitch spiral structure to study underwater low-frequency sound insulation and calculated the sound insulation effect using different parameters. Our results show that the fixed-pitch spiral structure has a better effect on underwater sound insulation and can achieve 3505 Hz–5355 Hz wide-band sound insulation. This spiral structure breaks through the limitation of appearance and can achieve sound insulation at the frequency of interest without changing the outer profile of the structure. Accordingly, it has potential in underwater low-frequency sound insulation applications.

Adaptive Federated IMM Filter for AUV Integrated Navigation Systems.

High accuracy and reliable navigation in the underwater environment is very critical for the operations of autonomous underwater vehicles (AUVs). This paper proposes an adaptive federated interacting multiple model (IMM) filter, which combines adaptive federated filter and IMM algorithm for AUV in complex underwater environments. Based on the performance of each local system, the information sharing coefficient of the adaptive federated IMM filter is adaptively determined. Meanwhile, the adaptive federated IMM filter designs different models for each local system. When the external disturbances change, the model of each local system can switch in real-time. Furthermore, an AUV integrated navigation system model is constructed, which includes the dynamic model of the system error and the measurement models of strapdown inertial navigation system/Doppler velocity log (SINS/DVL) and SINS/terrain aided navigation (SINS/TAN). The integrated navigation experiments demonstrate that the proposed filter can dramatically improve the accuracy and reliability of the integrated navigation system. Additionally, it has obvious advantages compared with the federated Kalman filter and the adaptive federated Kalman filter.

Active Sonar Target Classification with Power-Normalized Cepstral Coefficients and Convolutional Neural Network

Detection and classification of unidentified underwater targets maneuvering in complex underwater environments are critical for active sonar systems. In previous studies, many detection methods were applied to separate targets from the clutter using signals that exceed a preset threshold determined by the sonar console operator. This is because the high signal-to-noise ratio target has enough feature vector components to separate. However, in a real environment, the signal-to-noise ratio of the received target does not always exceed the threshold. Therefore, a target detection algorithm for various target signal-to-noise ratio environments is required; strong clutter energy can lead to false detection, while weak target signals reduce the probability of detection. It also uses long pulse repetition intervals for long-range detection and high ambient noise, requiring classification processing for each ping without accumulating pings. In this study, a target classification algorithm is proposed that can be applied to signals in real underwater environments above the noise level without a threshold set by the sonar console operator, and the classification performance of the algorithm is verified. The active sonar for long-range target detection has low-resolution data; thus, feature vector extraction algorithms are required. Feature vectors are extracted from the experimental data using Power-Normalized Cepstral Coefficients for target classification. Feature vectors are also extracted with Mel-Frequency Cepstral Coefficients and compared with the proposed algorithm. A convolutional neural network was employed as the classifier. In addition, the proposed algorithm is to be compared with the result of target classification using a spectrogram and convolutional neural network. Experimental data were obtained using a hull-mounted active sonar system operating on a Korean naval ship in the East Sea of South Korea and a real maneuvering underwater target. From the experimental data with 29 pings, we extracted 361 target and 3351 clutter data. It is difficult to collect real underwater target data from the real sea environment. Therefore, the number of target data was increased using the data augmentation technique. Eighty percent of the data was used for training and the rest was used for testing. Accuracy value curves and classification rate tables are presented for performance analysis and discussion. Results showed that the proposed algorithm has a higher classification rate than Mel-Frequency Cepstral Coefficients without affecting the target classification by the signal level. Additionally, the obtained results showed that target classification is possible within one ping data without any ping accumulation.

A Novel Efficient Passive Spatial Orientation Detection Method of UMT Enabled by ISB

The passive detection and direction-of-arrival (DOA) estimation problem is of great importance in many underwater applications such as target reconnaissance and data collection. In this paper, an Efficient Correlation-based Orientation Detection (ECOD) method is proposed to achieve high efficiency. Without high computational complexity in any Transform Domain, the time consumption of ECOD is largely reduced, which is especially critical for underwater intrusion detection, territorial waters protection, and many other real-time underwater applications. To achieve good invisibility, we design an intelligent submerged buoy (ISB) structure, which consists of six embedded hydrophones and an in situ electronic control unit (IECU). As a supplement to solutions against complex underwater environments, a hybrid ECOD method is also developed by involving the cooperation from underwater distributed sensor networks. To be specific, when high SNR signals are not recorded by a single ISB node, other distributed sensors are scheduled to assist in cooperative sensing. Simulation experiments demonstrate the efficiency of the ECOD method in passive 3D spatial orientation of underwater acoustic target and show that the ECOD method has a better performance in time consumption compared with general DOA algorithms.

Huber’s M-Estimation-Based Cubature Kalman Filter for an INS/DVL Integrated System

To deal with the problems of outliers and nonlinearity in the complex underwater environment, a Huber’s M-estimation-based cubature Kalman filter (CKF) is proposed for an inertial navigation system (INS)/Doppler velocity log (DVL) integrated system. First, a loosely coupled INS/DVL integrated system is designed, and the nonlinear system model is established in the case of big misalignment angle. Then, Huber’s M-estimation is introduced for robust estimation to resist outliers. Meanwhile, the CKF is focused to handle the nonlinearity of the state equation. Finally, simulation and the vehicle test are conducted to evaluate the effectiveness of the proposed method. Results show that the proposed method outperforms the conventional Kalman filter (KF) and outlier detection-based robust cubature Kalman filter (RCKF) in terms of navigation accuracy in the complex underwater environment.

Multi-autonomous underwater vehicle formation control and cluster search using a fusion control strategy at complex underwater environment

We introduce a multiple autonomous underwater vehicle (multi-AUV) control strategy to improve the cooperative operation ability in formation control and cluster search operations and to minimize the likelihood of collisions and interference with teammates and obstacles under various complex working conditions. Our approach uses a fusion control strategy with a redistribution mechanism (RM) based on virtual structure and leader–follower formation characteristics to improve the rapidity, stability, and accuracy of the multi-AUVs. The RM improved the rapidity and uniqueness of the random distribution formation by addressing over- and under-allocation. The combination of two (26-point and 8-point) path updating methods based on a hybrid bio-inspired self-organizing map algorithm considerably enhanced the performance of the multi-AUV system. Simulation tests and performance analyses indicated that the 8-point path updating method performed better in terms of rapidity; however, the 26-point path updating method was superior in integrity. The hunting performance of the fusion-controlled multi-AUV formation, was then analyzed to verify the feasibility of real-time formations under dynamic and static mixing of multi-obstacles by considering shape irregularities and the random distribution of multi-obstacles in three-dimensional underwater environments.

Underwater image enhancement framework and its application on an autonomous underwater vehicle platform

Underwater imaging has been increasingly employed in vision-based marine research. However, the inappropriate installation of a light source and the complex underwater environment will result in the uneven illumination and overexposure on the captured images. To address these issues, an underwater image enhancement framework for autonomous underwater vehicles platform is proposed, which consists of underwater light source optimization and illumination nonuniformity correction. The light source optimization method improves the imaging quality by computing an appropriate angle of the light casting. In this way, the center of the field of view is always well lit. In addition, an adaptive filter-based illumination correction algorithm is proposed to solve the uneven illumination caused by the artificial light source. During this process, image block segmentation and the measure of image enhancement index are applied to improve the adaptability and reduce the calculation errors of the filter parameters. A dataset with real underwater images collected under different natural conditions has been built and tested. The experimental results indicate that the proposed method is more adaptive and effective than the typical methods.

Exploring Underwater Target Detection Algorithm Based on Improved SSD

As the in-depth exploration of oceans continues, the accurate and rapid detection of fish, bionics and other intelligent bodies in an underwater environment is more and more important for improving an underwater defense system. Because of the low accuracy and poor real-time performance of target detection in the complex underwater environment, we propose a target detection algorithm based on the improved SSD. We use the ResNet convolution neural network instead of the VGG convolution neural network of the SSD as the basic network for target detection. In the basic network, the depthwise-separated deformable convolution module proposed in this paper is used to extract the features of an underwater target so as to improve the target detection accuracy and speed in the complex underwater environment. It mainly fuses the depthwise separable convolution when the deformable convolution acquires the offset of a convolution core, thus reducing the number of parameters and achieving the purposes of increasing the speed of the convolution neural network and enhancing its robustness through sparse representation. The experimental results show that, compared with the SSD detection model that uses the ResNet convolution neural network as the basic network, the improved SSD detection model that uses the depthwise-separated deformable convolution module improves the accuracy of underwater target detection by 11 percentage points and reduces the detection time by 3 ms, thus validating the effectiveness of the algorithm proposed in the paper.

A Support Vector Regression-Based Integrated Navigation Method for Underwater Vehicles

When Doppler velocity log (DVL) works in a complex underwater environment, it has the possibility of malfunction at any time, which will affect the positioning accuracy of underwater integrated navigation system (INS). In this work, the INS/DVL integrated navigation system model is established to deal with DVL malfunctions, and the support vector regression (SVR) algorithm is used to establish the velocity regression prediction model of DVL. An optimized grid search-genetic algorithm is used to select the best parameters of SVR. Simulations are designed to compare the results of SVR prediction model and isolating DVL during DVL failure. The semi-physical experiment is carried out to verify the validity and applicability of DVL velocity prediction model. The experimental results show that the INS/DVL integrated navigation system with the proposed model based on SVR performs better than the original integrated navigation system during DVL malfunction.

Learning-Based Dark and Blurred Underwater Image Restoration

Underwater image processing is a difficult subtopic in the field of computer vision due to the complex underwater environment. Since the light is absorbed and scattered, underwater images have many distortions such as underexposure, blurriness, and color cast. The poor quality hinders subsequent processing such as image classification, object detection, or segmentation. In this paper, we propose a method to collect underwater image pairs by placing two tanks in front of the camera. Due to the high-quality training data, the proposed restoration algorithm based on deep learning achieves inspiring results for underwater images taken in a low-light environment. The proposed method solves two of the most challenging problems for underwater image: darkness and fuzziness. The experimental results show that the proposed method surpasses most other methods.

Multi-autonomous underwater vehicles collaboratively search for intelligent targets in an unknown environment in the presence of interception

We present a bionic neural wave network that uses multiple autonomous underwater vehicles to search and acquire intelligent targets in an unknown underwater environment. The neuron pheromone content is arranged according to neural wave diffusion and layer-by-layer energy attenuation, when underwater mesh space based on neural wave diffusion theory was established that the neuron nodes in the neural network structure correspond to obstacles, autonomous underwater vehicles, and targets in the environment. In order to solve the problems of over-allocation and under-allocation of the multi-autonomous underwater vehicles system during the cooperative capture of targets, a redistribution mechanism based on the improved self-organizing map algorithm is implemented and directed to rationalize task distribution. Two different taboo search methods are employed to update the autonomous underwater vehicle path in real time, and the polynomial coefficient solution method is used to fit partial path data. So that the autonomous underwater vehicle trajectory can be obtained and an interceptor position coordinate can be predicted. An auxiliary autonomous underwater vehicle is aimed to replace the intercepted autonomous underwater vehicle and the matching capture points are tracked to ensure the completion of the task so that the full range of hunting targets is identified. In order to simulate an unknown complex underwater environment, obstacles are randomly arranged around the target, the location information of the obstacle, and the target is unknown and unpredictable. Four simulation experiments were performed to verify the accuracy and efficiency of the algorithm under unknown environment. The results show that this algorithm can improve the path update average efficiency by 66% compared with other algorithms. Obviously, this algorithm is reasonable and effective.