Research On Binocular Positioning System For Deep-sea Underwater Operation

Underwater binocular vision positioning is one of the core technologies of deep-sea intelligent precision operation.It plays a key role in guiding the precise grasping of deep-sea hydraulic manipulators and is a research hotspot in the process of developing China into a maritime power.However,the application of binocular stereo vision technology in deep-sea water environment is full of great challenges.First of all,there is a refraction phenomenon in operations under deep-sea water,and the refractive index of sea water fluctuates with changes in water depth,salinity,and temperature,which affects calibration errors and reduces positioning accuracy.Secondly,the shooting environment in deep-sea is harsh,plankton,undercurrent,flocculent and underwater operation jitter will affect the image quality,interfere with the extraction of image feature information,and reduce the accuracy of image matching.With the support of the sub-project "Development of Intelligent and Precise Operation Modules" of "Chinese Academy of Sciences Strategic Pilot Technology Special Project(Class A)",this thesis has carried out research work around the above two aspects,and studies underwater refraction model and camera calibration,underwater image enhancement and binocular stereo matching algorithm.The binocular positioning system is built,and the positioning experiment is carried out in the simulated deep-sea operation environment.The experimental results meet the requirements of deep-sea accurate operation.The specific research content of this paper is as follows:1.Research the principle of underwater binocular vision positioning and camera calibration technology.Aiming at the refraction problem in underwater operations,a refraction model is established to analyze the difference between the camera imaging in the air and underwater environments.According to the corresponding relationship between the threedimensional coordinates of underwater and air images,a mathematical model for transforming underwater images to air images is constructed,and the underwater images are corrected,which is verified by experiments.Finally,combined with the principle of binocular triangulation,a method based on mark points assisted calibration is proposed,which solves theproblem of calibration error caused by the dynamic changes of medium parameters in deep-sea water.2.Research on underwater image enhancement technology.In view of the characteristics of underwater image noise and blurring,bilateral filtering and defogging algorithm based on dark channel processing are adopted;Aiming at the characteristics of underwater image color shift and low contrast,first uses white balance and histogram equalization processing to get two images as the input of fusion image;then defines fusion weight to calculate normalized weight image;finally,the input image and the normalized weight map are multi-scale decomposed and fused based on the pyramid to obtain the fusion-enhanced image.3.The binocular stereo matching algorithm is studied,and corresponding solutions are proposed for the problems in the matching process.Aiming at the particularity of underwater operations,a semi-global matching algorithm based on Census transform matching cost is adopted,and the experimental results show that it is feasible.Finally,a binocular positioning experiment is carried out in a simulated deep-sea environment,and the object depth information is accurately measured according to the matched disparity map.The experimental data shows that it has a high accuracy in the working range of 1m?1.5m.