Based on analysis of state-of-the-art research investigating target detection and recognition in turbid waters, and aiming to solve the problems encountered during target detection and the unique influences of turbidity areas, in this review, the main problem is divided into two areas: image degradation caused by the unique conditions of turbid water, and target recognition. Existing target recognition methods are divided into three modules: target detection based on deep learning methods, underwater image restoration and enhancement approaches, and underwater image processing methods based on polarization imaging technology and scattering. The relevant research results are analyzed in detail, and methods regarding image processing, target detection, and recognition in turbid water, and relevant datasets are summarized. The main scenarios in which underwater target detection and recognition technology are applied are listed, and the key problems that exist in the current technology are identified. Solutions and development directions are discussed. This work provides a reference for engineering tasks in underwater turbid areas and an outlook on the development of underwater intelligent sensing technology in the future.

In the field of underwater vision, image matching between the main two sensors (sonar and optical camera) has always been a challenging problem. The independent imaging mechanism of the two determines the modalities of the image, and the local features of the images under various modalities are significantly different, which makes the general matching method based on the optical image invalid. In order to make full use of underwater acoustic and optical images, and promote the development of multisensor information fusion (MSIF) technology, this letter proposes to apply an image attribute transfer algorithm and advanced local feature descriptor to solve the problem of underwater acousto-optic image matching. We utilize real and simulated underwater images for testing; experimental results show that our proposed method could effectively preprocess these multimodal images to obtain an accurate matching result, thus providing a new solution for the underwater multisensor image matching task.

Image feature matching is essential in many computer vision applications, and the foundation of matching is feature detection, which is a crucial feature quantification process. This manuscript focused on detecting more features from underwater acoustic imageries for further ocean engineering applications of autonomous underwater vehicles (AUVs). Currently, the mainstream feature detection operators are developed for optical images, and there is not yet a feature detector oriented to underwater acoustic imagery. To better analyze the suitability of existing feature detectors for acoustic imagery and develop an operator that can robustly detect feature points in underwater imageries in the future, this manuscript compared the performance of well-established handcrafted feature detectors and that of the increasingly popular deep-learning-based detectors to fill the gap in the literature. The datasets tested are from the most commonly used side-scan sonars (SSSs) and forward-looking sonars (FLSs). Additionally, the detection idea of these detectors on the acoustic imagery phase congruency (PC) layer was innovatively proposed with the aim of finding a solution that balances detection accuracy and speed. The experimental results show that the ORB (Oriented FAST and Rotated BRIEF) and BRISK (Binary Robust Invariant Scalable Keypoints) detectors achieve the best overall performance, the FAST detector is the fastest, and the PC and Sobel layers are the most favorable for implementing feature detection.

This study considers the data processing for acoustic cameras and achieves the generation of high-quality acoustic panoramas through image mosaicking. Thanks to high-resolution imaging, acoustic cameras are increasingly popular in ocean engineering. However, their narrow detection field of view makes it challenging to intuitively perceive marine environments. Generating large panoramas through mosaicking is a good way to solve this problem. Due to limitations such as low resolution, low signal-to-noise ratio, weak textures, and nonlinear distortions in acoustic images, most classic mosaicking pipelines do not perform well. This study proposes an adaptive mosaicking framework for acoustic cameras that integrates image denoising, feature matching, and mosaicking modules. It can generate large-area panoramas from overlapping acoustic camera images without any assumptions regarding the scenes. The overall process consists of three main steps: first, introduce a self-supervised denoising strategy to preprocess acoustic images to effectively remove complex noise; second, use a detector-free paradigm to achieve feature matching between adjacent acoustic images. This paradigm matches dense pixels in the high-level structure of images rather than relying on isolated geometric features, addressing the matching challenges in weak-texture areas. Third, design a mosaicking approach based on matching results to generate acoustic panoramas. This framework has been verified experimentally, and the results show that it can robustly and effectively mosaic acoustic images, providing a novel reference and solution for underwater structures inspection in complex marine environments.

This article proposes an automatic sonar target shadow extraction approach based on a single acoustic camera to enable efficient three-dimensional (3D) perception of seafloor environments. The acoustic camera, a compact high-resolution forward-looking sonar, functions effectively in low-visibility conditions such as turbid and dark underwater environments. However, owing to the unique mechanism of sonar imaging, the captured two-dimensional sonar images generally lack elevation (height) information, thereby constraining rapid 3D seafloor perception. This study estimates target height by leveraging the cast shadow clues present in sonar images, in combination with the sensitivity of sonar to subtle variations in these visual clues. Using shadow information to solve seafloor target height is a classic paradigm in sonar visual applications, but traditional methods do not work well in complex seabed environments. Therefore, this study proposes a comprehensive shadow extraction approach based on deep learning to overcome these limitations. First, a learning-based extraction framework is employed to automatically identify cast shadows of sonar targets, addressing challenges posed by low signal-to-noise ratios and poor contrast. To improve the accuracy of shadow labeling, a sonar simulator is utilized to assist the annotation process by providing reference data for cast shadow identification. Then, by combining the extracted shadow data with the altitude difference derived from sonar, the seafloor target height is quickly resolved. Unlike previous 3D reconstruction pipelines, the proposed method does not rely on numerous overlapping views or complex elevation assumptions, enabling an efficient estimation of target height. Validation in an acoustic pool experiment shows that this method can reliably achieve efficient 3D seafloor perception, potentially serving as a pivotal technology to support autonomous operations of marine robots in unknown seafloor.

Underwater automatic target recognition (UATR) has been a challenging research topic in ocean engineering. Although deep learning brings opportunities for target recognition on land and in the air, underwater target recognition techniques based on deep learning have lagged due to sensor performance and the size of trainable data. This letter proposed a framework for learning the visual representation of underwater acoustic imageries, which takes a transformer-based style transfer model as the main body. It could replace the low-level texture features of optical images with the visual features of underwater acoustic imageries while preserving their raw high-level semantic content. The proposed framework could fully use the rich optical image dataset to generate a pseudo-acoustic image dataset and use it as the initial sample to train the underwater acoustic target recognition model. The experiments select the dual-frequency identification sonar (DIDSON) and adaptive resolution imaging sonar (ARIS) as the underwater acoustic data source and also take fish, the most common marine creature, as the research subject. Experimental results show that the proposed method could generate high-quality and high-fidelity pseudo-acoustic samples, achieve the purpose of acoustic data augmentation and provide support for the underwater acoustic-optical images domain transfer research.