Research On Underwater Pipeline Detection Algorithm Based On Side-scan Sonar

With the development of society and technology,land resources will gradually be exploited,and people will begin to focus on the exploitation of marine resources.In the21 st century,the pace of marine development has gradually accelerated.We are a maritime country with a long coastline,and becoming a maritime power is also the mission of our Chinese sons and daughters.For decades,in order to make full use of marine resources,countries have laid a large number of energy transmission pipelines and communication cables in the oceans.In order to ensure the normal operation of submarine pipelines,robots need to be used to regularly inspect submarine cables.Side scan sonar has become the main equipment for marine pipeline detection,through image processing scientific research to detect the location and direction of pipelines,this paper starts from the side scan sonar pipeline image,research the detection algorithm of underwater pipeline,the content is as follows:Firstly,the composition structure and imaging principle of side-scan sonar are studied.Firstly,we understand the mechanical and electronic components of side-scan sonar design,analyze its imaging principle and the reasons affecting the imaging quality of sonar images.Understand sonar packet composition and experiment with sonar datasets.A side-scan sonar noise model is established,and denoising methods for spatial domain,transformation domain and image morphology are proposed,and experimental verification shows that the detection effect of the denoising method based on the transformation domain is better than that of others.Secondly,the methods of fast line segment detection,Hough,random consistency detection and clustering algorithms on sonar pipeline detection are studied,and several types of algorithms are analyzed for pipeline detection,and an improved line feature detection algorithm is proposed to meet the real-time characteristics of detection.The algorithm is verified in the lake test,and the results show that the algorithm detects the pipeline target at scientific research,and the detection and tracking distance of the AUV from the pipeline is very small.Finally,based on the depth characteristics of sonar,the pipeline detection of sonar images is studied,and the Yolo-v3 network is proposed to detect underwater targets in side-scan sonar images.The deep network can combine feature extraction and mission objectives to optimize network parameters,and then automatically extract effective information of complex high-dimensional data,which greatly improves the detection and recognition accuracy of underwater targets and has good practical application value.