Research And Application Of Binocular Vision 3D Measurement Technology

Computer vision technology has been proposed since the 1980 s and has been developed for decades.It has been widely used in a variety of cutting-edge technologies.Binocular stereo vision,as a branch of computer vision,is often used to solve problems such as target measurement,target positioning,and 3D reconstruction.The nuclear reactor fuel assembly is the core component of the energy generated by the nuclear energy system.The accurate measurement of the fuel assembly is the guarantee for the safe and stable operation of the entire nuclear energy system.Traditional contact measurement,slow speed,high risk and high cost.Binocular vision has the advantages of non-contact,low cost,fast measurement speed,and the potential for non-contact measurement of fuel assemblies.In this paper,the various techniques of binocular vision 3D measurement are deeply studied,and the binocular vision system is designed.The system is applied to the measurement of fuel assembly size.In terms of binocular vision calibration,this paper firstly studies the single camera imaging model,the classic single camera calibration method,and the binocular camera calibration principle.On this basis,a binocular vision system was built,and the system was calibrated with a solid circular target calibration plate.After repeated experiments,the calibration accuracy of the system was stable at around 0.06 pixels.In the aspect of image feature extraction and stereo matching,this paper studies and analyzes the basic theory of stereo matching and the classical feature extraction algorithm.On this basis,this paper designs a NCC(normalized cross correlation)feature point matching method based on polar line correction.The fuel cell standard block image is used as the matching object.After repeated experiments,the method can effectively complete the stereo matching of key feature points.Finally,this paper studies and compares the common binocular vision threedimensional measurement methods,and proposes an improved binocular vision threedimensional measurement method by corner reconstruction.And it is applied to the standard block diagonal measurement of fuel assemblies.After repeated experiments,the measurement accuracy of the algorithm is 0.015 mm,which can meet the requirements of nuclear power plant measurement accuracy.The experiment proves binocular vision 3D measurement technology can be applied to solve the non-contact measurement problem of fuel assemblies.