Research On Simultaneous Localization And Map Construction Of Autonomous Underwater Vehicle

Simultaneous Localization And Mapping(SLAM)technology is a key technology for autonomous underwater robots to achieve underwater navigation,and the SLAM technology based on camera sensors is usually called: visual SLAM.Visual SLAM has the advantages of low cost,rich scenes and so on,so it is widely used.At present,compared to the application of visual SLAM technology in land and aerial scenes,its use in underwater environments is relatively few.In this paper,the vision SLAM system with front-end improvement and point line fusion is used to improve the positioning accuracy and map building effect in underwater environment.The main research contents are as follows:First of all,aiming at the problem that the visual slam is facing the change of light propagation medium when it is used in the underwater environment,the camera sensors used in it are analyzed and screened.The coordinate mapping relationship from the world coordinate system to the pixel coordinate system is derived.Zhang's calibration method is used to calibrate the stereo camera which is suitable for underwater use.The camera calibration experiments in the underwater environment and the land environment are carried out respectively.The accuracy of the calibration results of the underwater camera is compared and analyzed,and the problem of changing the internal parameters of the camera in underwater use is solved.Secondly,the front end of visual SLAM is improved to solve the problems of dim and fuzzy scene when it is used in underwater environment.The image enhancement module is added,and the improved histogram equalization algorithm is used to process the input image of the sensor.Through image evaluation index calculation,feature point extraction experiment and underwater experiment,it is proved that the image enhancement module has obvious improvement effect.Then,aiming at the problems of single feature and insufficient feature when visual SLAM is used in underwater environment,a visual SLAM system framework of point-line feature fusion is proposed.The ORB point feature and LSD line feature are extracted from the enhanced image,and the binary descriptors of point-line feature are calculated.In order to achieve better matching effect,quadtree algorithm is used to homogenize point features.Theerror model of reprojection and Jacobian matrix analytic expression of point line feature are derived.At last,aiming at the problems of structural feature change when visual SLAM is used in underwater environment,the dictionary library of loop detection module is trained and updated based on a large number of underwater image datasets.The improved point-line fusion pline-slam system is proved to have higher positioning accuracy under the KITTI datasets.At the same time,it carries on the underwater ROV and carries on the underwater experiment.The results show that the improved system has better mapping effect.