Font Size: a A A

Construction Of Visual Positioning And Simulation Platform Of Hot Cell Manipulator Based On Virtual Reality Technology

Posted on:2021-02-19Degree:MasterType:Thesis
Country:ChinaCandidate:P GuoFull Text:PDF
GTID:2512306755954749Subject:Mechanical engineering
Abstract/Summary:PDF Full Text Request
As a major equipment for the maintenance of nuclear power facilities,manipulators can be used to perform heavy and complex tasks in a nuclear environment hot cell inaccessible to humans.However,domestic hot cell manipulator operators only use the naked eye to observe the position of the work object through the narrow radiation-proof window during the operation,which is likely to cause safety hazards.Therefore,this paper carries out research on the visual positioning of power manipulators based on virtual reality(VR)technology in a nuclear environment,and the results of machine vision positioning was used to build a VR simulation platform to assist operators in observation and operations.Firstly,The operation process of the hot cell manipulator in the current domestic nuclear environment was investigated.Aiming at the interference of nuclear radiation high-energy particles on the precision sensing instrument,a overall plan that radiation-proof binocular camera for the front-end hot cell data collection and the outdoor host computer for the backend data processing is proposed.Secondly,the median filter algorithm is used to preprocess the image denoising for the radiation interference problem of the collected image data.Then an improved Yolo.v4-tiny hot cell manipulator object detection algorithm based on deep learning was designed,which significantly improved the speed and effect of object detection compared with the classic algorithm.After the camera is calibrated by Zhang Zhengyou celibration method,a deep learning object detection and BM binocular stereo matching fusion algorithm is proposed,which extends the two-dimensional information to three-dimensional,and realizes the spatial positioning of the target.Aiming at the missing detection objects and abnormal data caused by the holes in the depth map,a median and context comparison filtering script was developed.Finally,a VR simulation platform was built based on Unity3 D to realize the frontend and back-end data communication and real-time reconstruction of job objects.And the system platform was tested to verify the correctness and feasibility of the research.In this paper,the end-to-end spatial positioning and VR simulation of hot cell work objects in a nuclear environment was realized,thereby assisting operators in remote operations.The development of nuclear industry intelligent manipulators was provided ideas by its high-precision real-time positioning feature.
Keywords/Search Tags:Hot cell manipulator, Deep learning, Object detection, Visual positioning, Virtual reality
PDF Full Text Request
Related items