Submarine Pipeline Visual Detection And Target Positioning System For Underwater Autonomous Robots

The bottom of the ocean is a treasure trove of natural mineral resources.In recent years,humans have continuously exploited marine oil and gas resources and laid a large number of submarine pipelines on the seabed for transportation.As of the end of2009,the oil and gas transported by submarine pipelines accounted for more than20% of the national oil and gas supply,so the protection of submarine pipelines has also been put on the agenda of my country’s legislation.During daily pipeline operation,regular inspection of subsea pipelines is essential.Usually,the detection and maintenance of submarine pipelines are completed by underwater robot inspections.Therefore,the development of a submarine pipeline detection and positioning system for underwater robot inspection is of great significance for ensuring resource transportation and protecting the marine environment.In this thesis,the optical camera is used to obtain image information as a data source,and the detection and identification of the surface of the submarine pipeline and the positioning algorithm under the binocular vision system are run on the embedded platform,and a system that provides the direction of the pipeline and the location of the fault on the pipeline surface is designed to assist Inspection operations of underwater autonomous robots.Aiming at the problem of pipeline and fault detection,firstly,in the underwater image preprocessing process,this thesis adopts an improved Gaussian filtering method based on the gray value and distance of pixels,which can effectively reduce image noise and improve the signal-to-noise ratio compared with traditional Gaussian filtering.6.1d B;then use Gray World combined with the color compensation algorithm to improve the color cast of the image,and compare the renderings to find that the algorithm can restore the underwater image more realistically;further use the image enhancement algorithm to obtain a subsea pipeline image with strong contrast.Secondly,in the identification process of the submarine pipeline,according to the characteristics of the pipeline,this thesis proposes an improved adaptive Canny algorithm and an improved Hough algorithm,which can detect two edge lines close to the pipeline in the submarine pipeline image,and the recognition rate of the system test up to 94%.Finally,using morphological and histogram detection techniques,the location of the fault in the area where the pipeline is located in the image can be accurately located.Aiming at the problem of pipeline navigation,this thesis proposes a laser line-assisted binocular vision positioning method.This method extracts the feature points where the laser lines intersect on the pipeline surface for matching,which reduces the complex calculation of pipeline linear features.The obtained multiple 3D point coordinates are estimated by the cylinder fitting method to estimate the pipeline pose.Aiming at the fault point navigation problem,this thesis adopts the matching method to carry out mean fitting on the three-dimensional points.Finally,the feasibility of the system is verified through system tests in laboratory water tanks and outdoor pool scenarios.Further,the positioning accuracy of the vision system is verified to be no more than 2mm at a visual distance of 1m by using a standard block with an accuracy of <0.1mm,and the possible errors are analyzed based on the measurement results.