Research on Detection Method for the Leakage of Underwater Pipeline by YOLOv3
Xinhua Zhao, Xue Wang, Zeshuai Du
Abstract
With the increasing demand for marine development, underwater robots have become more and more widely used in the fields of underwater data monitoring, seabed detection, marine target acquisition and recognition, especially in the detection of oil leaking points in underwater pipelines. In this paper, a method for detecting the oil spill point of the underwater pipeline based on YOLOv3 is proposed. Aiming at the problems of image distortion, blur and low contrast, the method of Gauss filtering, brightness enhancement, sharpening, and histogram equalization are used to improve the image quality. The enhanced pipeline images are brought into YOLOv3 as a training set for training: the input images and their labels are sent into the darknet-53 network, then the prediction result of the network is processed to get the detection target according to logical regression, and the object classification and location are completed in one step, which improves the efficiency of detection. The experimental results show that the network can detect pipeline vulnerabilities quickly, and the accuracy is high, and the missed detection rate is low.