Combining artificial intelligence and laboratory experiments to explore behavior process of sunken and submerged oil: A typical oil drift and diffusion detection technology

Abstract

This paper aims to use laboratory experiments to study the behavior process of sunken and submerged oil under different breaking waves and then combines the experimental results with deep learning technology to establish a typical oil drift and diffusion model. Using the image data of real oil pollution moving under breaking wave energy in the laboratory, this research established and compared the ‘You Only Look Once v4 model’, ‘CenterNet model’, ‘You Only Look Once v3 model’, and ‘Single Shot MultiBox Detector model’. The results show that the average detection speed values of ‘You Only Look Once v4’, ‘You Only Look Once v3’, ‘CenterNet’, and ‘Single Shot MultiBox Detector’ are 42.47 frames per second, 26.80 frames per second, 35.37 frames per second, and 38.90 frames per second, respectively. In terms of detection accuracy, the ‘You Only Look Once v4’ algorithm exceeds the other three algorithms by 10.59%, 22.62%, and 31.69%. Considering its detection speed and detection accuracy, the ‘You Only Look Once v4’ algorithm has strong robustness to changes in the shape and movement of sunken and submerged oil. In addition, by using ‘You Only Look Once v4’ model for oil detection, it is found that the submergence time of oil droplet for 10 cm, 15 cm, 20 cm, 25 cm and 30 cm breaking waves is 2.32 s, 2.52 s, 2.62 s, 3.20 s, 7.12 s, respectively. The speed of sinking to the deepest point is 0.208 m/s, 0.222 m/s, 0.212 m/s, 0.359 m/s, 0.303 m/s.