Underwater target detection exhibits extensive applications in marine target exploration and marine environmental monitoring. However, conventional images of underwater targets present challenges including blurred contour information, complex environmental conditions, and pronounced scattering effects. In this work, an underwater target detection method based on YOLOv10 is designed, and the detection performance is compared with the YOLOv5 model. Experimental results demonstrate that the YOLOv10 model has a mAP50 of 85.6% on the URPC 2020 dataset, improving the mAP50 by 1.2% than that of YOLOv5. This model exhibits high detection accuracy and high proceeding speed, which provides a promising support for precise and fast underwater target detection.

Brillouin optical time-domain analyzer (BOTDA) is used to monitor the temperature and strain along a fiber. So far, neural network and machine learning methods have been successfully applied for temperature extraction. But for different frequency scanning steps, different networks should be designed and trained. Here, a BOTDA assisted by kernel extreme learning machine (K-ELM) with high generalization is proposed and experimentally demonstrated. By utilizing K-ELM, the raw Brillouin gain spectra measured from BOTDA system are classified into different temperature classes. The performance of K-ELM is investigated both in simulation and experiment under different cases of signal-to-noise ratios, pump pulse widths, and frequency scanning steps. Compared with curve fitting methods, the K-ELM algorithm has better measurement accuracy of 0.3 and it realizes great improvement of the processing speed over 120 times. The ultrafast processing speed, high accuracy and universality make K-ELM become a highly competitive candidate for the high-speed BOTDA sensing system.

The fiber-optic hydrophone (FOH) provides new possibilities for underwater acoustic detection and surveillance, particularly when integrated with autonomous underwater vehicles (AUVs). Its lightweight, high sensitivity, and ease of multiplexing are significant advantages. Miniaturizing the FOH and incorporating multiple units into a thin towed array suitable for AUVs, however, presents a challenge. In this study, we present the design and implementation of a 48-element thin FOH towed array, boasting an outer diameter of only 20 mm. The average pressure sensitivity of the FOH is approximately -140.2 dB re rad/μPa, with a fluctuation of less than ± 1 dB across a frequency range of 20 Hz – 4 kHz. This wide operating frequency range and high sensitivity align well with the demands of AUVs for underwater acoustic detection and surveillance. The average minimum detectable pressure of the array meets practical application requirements, even when compared to ambient sea noise. Noise characterization was assessed through towing tests in real-world environments. Results indicate that flow noise, dominant in the low-frequency band (< 200 Hz), is unlikely to be a limiting factor in practical applications, as the design frequencies of the array are located outside this range. The directional of arrival (DOA) estimation of the fixed and moving sources was successfully demonstrated. Bearings obtained using a conventional beamformer closely matched the true values, confirming the feasibility of the thin FOH towed array for practical DOA estimation. These findings are instrumental for the future implementation of thin FOH towed arrays on AUVs.

This work investigates the problem of estimating directions of arrival (DOAs) of coherent sources without the left–right ambiguity by acoustic vector sensor linear arrays. Toeplitz matrices obtained from vector sensor cross covariance matrices are summed to obtain a new matrix with vector sensor directivities as weights. In this way, the proposed method restores the rank of the equivalent signal subspace and eliminates 2 π ambiguous DOA information of sources. A robust technique is also developed to improve its performance degradation because of the velocity sensor gain uncertainty during unit conversion, such as imprecise measurements of the sensor sensitivities and the product of the water density and sound speed in the water. The methods can be implemented as a preprocessing stage before applying beamformers, such as multiple signal classification and MVDR. Theoretical analyses show that comparing with existing preprocessing techniques, the proposed methods can fully resolve the coherent sources without the left–right ambiguity while keeping the effective array aperture, and even applicable to the case of closely spaced coherent sources. Their effectiveness and performances are conducted via simulations.

Background normalization algorithms attempt to suppress the ambient and self-noise during the measurements of sonar, which enhance the detection performance and the display effect of weak signals. Conventional background normalization methods are usually sensitive to the accuracy of prior set filtering interval and threshold, while significant noise is still detected in low frequency. In this paper, an improved background normalization algorithm is proposed by thresholding the processing interval between several local peak values and local valley values. Compared to the existing scenarios, the proposed approach automatically calculates the filtering interval and threshold, with substantial resilience to the noise level in low frequency. Experimental results illustrate the effectiveness of our algorithm.

Signal detection and format recognition are critical and challenging tasks across civil and military sectors. However, they often encounter signal truncation issues during online signal processing, resulting in inaccurate predictions due to incomplete features. To address this issue, we herein propose the signal detection mask long short-term memory (SDM-LSTM) network module. The SDM-LSTM module facilitates seamless transmission of sequential signal features among blocks, even in scenarios facing signal truncation. Our approach involves organizing online-acquired signal samples into short blocks and conducting signal feature extraction and prediction for each block. This methodology ensures that each block retains the cumulative features from preceding blocks, thereby maintaining the continuity of signal features. Consequently, signal truncation no longer induces feature incompleteness. The proposed model, leveraging the SDM-LSTM module, offers predictions for each block, aligning with the sequence-to-sequence prediction paradigm. The simulations and experimental results validate the efficacy of our model in addressing signal truncation and facilitating online signal processing.