A real-time detection method for underwater pipeline in side scan sonar images based on semantic segmentation
To fill the research gap in automatic diagnosis of underwater pipeline burial status using SBP images and improve the automation level of underwater pipeline inspection,a complete set of automatic diagnosis methods and processes for underwater pipeline burial status has been provided.Firstly,efficient data preprocessing methods were used to accurately restore the true information of pipelines.Secondly,accurate extraction of seabed lines was achieved based on Frangi filter enhancement technology.Then,deep learning technology was used to achieve high reliability detection of pipeline targets.Finally,criteria for determining the burial status of pipelines was provided,and the burial status of pipelines was automatically determined using the positional relationship between pipeline detection results and the seabed.Experiments were conducted using measured data from various types of shallow layer profilers,and the results showed that the detection accuracy of underwater pipelines can reach a Recall of 0.952 and a mAP of 0.962.Based on the target detection,accurate diagnosis of pipeline burial status can be achieved.
国家科技期刊平台
NSTL
万方数据
