Multi-level Map Construction Of Unmanned Biosafety Mobile Laboratory With Heterogeneous Binocular Vision

In recent years,the outbreak of the African swine fever epidemic,the Ebola virus and anthrax terrorist attacks have caused the global biosafety issue to renew the concern of the whole world.In the treatment of biosafety issues,research aimed at rapid detection of highly pathogenic and lethal microorganisms on the spot is a hot spot.Therefore,biosafety mobile laboratories meeting the actual needs of the site have emerged.However,existing biosafety mobile laboratories have certain limitations,such as large size,poor mobility,and the safety of the experimental personnel can't be fundamentally guaranteed.With the development of the intelligent robot industry,this paper intends to carry out research on key technologies of intelligent robot-based unmanned biosafety mobile laboratories,focusing on the identification of multi-level maps of unmanned biosafety mobile laboratory instruments and equipment based on binocular vision system.The construction,precise positioning of experimental operations and other issues.The following research work was carried out.(1)Design and implementation of unmanned biosafety mobile laboratory and binocular vision system:the design of unmanned biosafety mobile laboratory adopts mechanical arm instead of human scheme,and plans the layout of unmanned biosafety laboratory;the heterogeneous double visual system uses a color plus black-and-white camera module to obtain color and black-and-white image information.The coordinate system transformation matrix of the binocular vision module is solved according to the motion pose of the robot arm to complete the calibration of the binocular vision module.(2)Multi-level map construction of unmanned biosafety mobile laboratory:According to the overhead image of the unmanned biosafety mobile laboratory,the primary map for path planning is established by using the envelope block instead of the instrument;in the experimental target area,through the ring shot a continuous number of point clouds are obtained,and a point cloud registration algorithm is used to construct a secondary map for accurately identifying instruments and equipment;an instrument device can be used to expand the three-dimensional model database,and a training device training set is trained to identify and retrieve the target instrument device.A three-level map for precision experiments was constructed using the models in the 3D model database.(3)Accurate experiments of instruments and equipment based on multi-level maps.Taking the experiment of the liquid-pipe multi-hole plate with the highest positioning accuracy as an example,the experimental error is analyzed,and the coordinates of the center of the hole in the perforated plate are accurately positioned by the distortion correction-circle contour extraction-feature matching method.The main innovations of this paper are as follows:1.Established a multi-level map based on different application scenarios of heterogeneous binocular vision systems.The multi-level map is divided into three levels,the primary map is the envelope block map,the heterogeneous binocular black and white image is used to sample the contour,the color image is used to supplement the color,the primary map is applied to the path planning,and the second level map is the point cloud map.The black and white image of the heterogeneous binocular calculates the depth information,the color image is accurately identified,the second level map is applied to the precise positioning instrument;the third level map is the three-dimensional model map,the black and white image in the heterogeneous binocular is used for the distortion correction,and the color image is used for the distortion correction.Feature screening,three-level maps for precision experiments.2.A rotation-defined iterative nearest point cloud registration algorithm RL-ICP(Rotating Limit-Iterative Closest Point)is proposed for 3D image stitching.It solves the problem that the common registration algorithm ICP initial matrix parameters are unknown,and it is easy to fall into the local optimal solution.RL-ICP defines the iterative European space threshold and rotation matrix weight,k-d tree and backtracking search to avoid the local optimal solution and improve the registration rate of the point cloud.3.A matching correction method(DC-Hough-SURF)is proposed for distortion correction,Hough contour extraction and SURF feature detection.It is used to solve the precise positioning problem of the experimental operation of the unmanned biosafety mobile laboratory,such as the dripping operation of the pipetting gun in the multi-well plate experiment.Distortion correction is used to solve the problem of correcting the image taken by the camera oblique angle due to the viewing angle blocking;Hough circular contour extraction is used to solve the mismatching problem in feature matching;SURF feature detection can solve the actual captured image and the top view image in the 3D model database.Feature matching problem.The positioning accuracy of the method can reach the positioning requirement of the pipetting drip plate with the highest accuracy of the experimental positioning.