Autonomous Underwater Vehicle Control for Fishnet Inspection in Turbid Water Environments

Fisheries are essential for the economic supply of proteins. Detecting damaged fishnets using autonomous underwater vehicles (AUVs) may be an efficient and safe solution for avoiding dangers to human divers. However, in turbid underwater environments, visibility is significantly degraded by floating particles that cause light attenuation, which is one of the main problems for accurate underwater inspection by optical cameras. To obtain clear images for net inspection, we propose an AUV pose control strategy for fish farming net inspection in turbid water, based on the mean gradient feature over the partial or entire image. To alleviate the laborious human process of setting the desired set-point for distance control, a convolutional neural network (CNN) is trained offline using a supervised learning method and combined with a controller. The proposed method can maintain a relatively constant relative pose with respect to a fishnet, which is sufficient to acquire clear net images in turbid water and check whether a part of the net is damaged or not. Experimental results in both swimming pools and real fish farm environments demonstrated the effectiveness of the proposed methods.

Automated summary

Fisheries are crucial for protein supply in the economy. Using autonomous underwater vehicles (AUVs) to detect damaged fishnets offers an efficient and safe solution in turbid underwater environments. This study proposes an AUV pose control strategy based on image mean gradient feature and a convolutional neural network (CNN) combined with a controller, enabling clear net images acquisition and damage inspection in turbid water.