Research On Key Technologies Of Machine Vision-based AUV Terminal Guidance System

AUV,namely Autonomous Underwater Vehicle,can autonomously complete the pre-set tasks with the energy battery and the control host carried by itself.However,AUV's working time and working scope are limited by the energy.With the development of AUV technology,the docking of AUV is now the research focus at home and abroad.This paper mainly works at the research on key technologies of the machine vision-based AUV terminal guidance system.Chapter 1 firstly describes the research background,then reviews on domestic and overseas research progress of underwater machine vision localization.On basis of existing research,this paper discusses the research contents and technical difficulties.Chapter 2 mainly discusses the research results of underwater image enhancement and target detection technologies.The quality of underwater imaging is poor because of the absorption and scattering phenomenon in water medium.Therefore,the underwater image must be enhanced to obtain the image features of interest.In this paper,the camera Auto-Exposure algorithm and the image adaptive threshold segmentation algorithm are studied.In addition,this paper uses the affine-invariant moments as the eigenvalue of the SVM classifier to identify underwater targets.Chapter 3 firstly analyzes the optical characteristics of the sea water,which is used to choose underwater lamp.In addition,the linear geometric model of camera imaging and the mechanism of visual guidance with monocular camera and single lamp are discussed.After building the docking platform,this paper has carried out the pool experiments,lake experiments and sea trials.On the basis of Chapter 3,Chapter 4 proposes two methods of 3D positioning of underwater target.In order to obtain accurate camera parameters,this paper firstly expounds the nonlinear geometric model of camera imaging and the camera calibration method.In this paper,two 3D positioning methods are proposed,including the monocular vision 3D positioning based on the known model and the the binocular vision 3D positioning.The experiment result shows that the above two methods both work well.In comparison,the positioning accuracy of binocular vision 3D positioning is higher,but the distance of the target can not be too small because of the limitation of the FOV.Chapter 5 uses a surface boat as the carrier to carry out the visual-based terminal guidance experiment in the laboratory pool.The experiment result shows that the method proposed in this paper can accurately guide the experiment platform to the target.Chapter 6 summarizes the research process and results,and then puts forward some suggestions for further research.