Search

The issue of marine litter has been an important concern for marine environmental protection for a long time, especially underwater litter. It is not only challenging to clean up, but its prolonged presence underwater can cause damage to marine ecosystems and biodiversity. This has led to underwater robots equipped with powerful visual detection algorithms becoming the mainstream alternative to human labor for cleaning up underwater litter. This study proposes an enhanced underwater litter detection algorithm, YOLOv7t-CEBC, based on YOLOv7-tiny, to assist underwater robots in target identification. The research introduces some modules tailored for marine litter detection within the model framework, addressing inter-class similarity and intra-class variability inherent in underwater waste while balancing detection precision and speed. Experimental results demonstrate that, on the Deep Plastic public dataset, YOLOv7t-CEBC achieves a detection accuracy (mAP) of 81.8%, markedly surpassing common object detection algorithms. Moreover, the detection frame rate reaches 118 FPS, meeting the operational requirements of underwater robots. The findings affirm that the enhanced YOLOv7t-CEBC network serves as a reliable tool for underwater debris detection, contributing to the maintenance of marine health. [ABSTRACT FROM AUTHOR]

Underwater image processing has always been a very challenging problem. Under the influence of environmental factors, underwater images are prone to some problems, such as colour cast, low visibility, and few edge details. Here, an image enhancement algorithm is proposed to improve image degradation mainly caused by the absorption of light. First, colour compensation and white balance algorithm are used to restore the natural appearance of the image. Then the improved dark channel prior (DCP) is used to improve the visibility and avoid blocking artifacts which appear in traditional DCP. Unsharp masking (USM) is applied to enhance the texture features of the DCP image. Finally, wavelet fusion is used to fuse the DCP image and DCP+USM image. The fusion algorithm not only further improves the visibility and texture features, but also reduces the noise of DCP+USM. Compared with other methods, quantitative analysis results show that the enhanced images have higher visibility, more details and edge information. [ABSTRACT FROM AUTHOR]

For a class of remotely operated vehicle (ROV) systems with thruster constraints, immeasurable states, and unknown nonlinearities, the trajectory tracking control problem was discussed in this paper. The unknown nonlinear functions were approximated by radial basis function (RBF) neural networks. An adaptive state observer based on neural networks was designed and the immeasurable states were estimated. Considering the problem of thruster saturation constraints, an auxiliary system for saturation compensation was designed and a saturation factor was constructed by the auxiliary system state. By applying the backstepping design method, an adaptive neural sliding mode trajectory tracking controller was developed, in which the saturation factor is contained in adaptive laws. It was proved that the uniformly ultimately bounded (UUB) of trajectory tracking errors can be obtained. Finally, the effectiveness of the proposed trajectory tracking control approach was checked by simulations. [ABSTRACT FROM AUTHOR]