Underwater Interdiction Net's Fine-Grained Defect Identification Algorithm

This paper proposes a fine-grained image classification algorithm for interdiction net's defect detection in nighttime and underwater scenarios. The algorithm is primarily applied to classify the different severity of fouling, corrosion and breakage conditions of underwater interdiction nets, avoiding situations where manual inspections are not possible due to unpreventable factors such as catastrophic weather and where real-time inspections cannot be performed manually. By analysing the image texture, shape, color space, roughness and Otsu thresholding, the proposed data-driven algorithm can further classify in real time the defective condition of underwater interdiction net that have been dynamically drifting and deforming in the deep sea.