Research On 3D Coordinate Positioning Method Of Deep Learning Based On Point Diffraction Interference

With the development of intelligent manufacturing industry,its requirements for measurement accuracy and measurement speed are increasing day by day,so point diffraction interference technology has been studied and developed in the field of three-dimensional coordinate positioning.Although there are many kinds of3 D coordinate positioning instruments,it is difficult to realize the requirements of small volume,easy to carry,no guide rail,high accuracy,fast speed and strong antiinterference ability.The traditional point diffraction interference 3D measurement system is easy to be interfered by the external environment,with low precision and slow operation speed.To solve these problems,this dissertation studied the 3D coordinate positioning method based on point diffraction interference,and analyzed the relationship between point light source coordinates and phase difference matrix.Finally,a deep learning 3D coordinate positioning method based on point diffraction interference is proposed,which realized good reliability,high precision and strong anti-interference ability.The main research contents of this dissertation are as follows:The research status of point diffraction interference and deep learning in the field of three-dimensional coordinate positioning is described.The deep learning 3D coordinate positioning measurement optical path model based on point diffraction interference is described.Meanwhile,the mathematical principle and flow of measurement are introduced.The possibility of applying deep learning technology in 3D coordinate positioning was analyzed deeply.Based on the traditional neural network model,a deep learning 3D coordinate positioning neural network was proposed for 3D coordinate positioning reconstruction.According to the requirements of actual system parameters and core distance,a large number of phase difference matrices were randomly obtained by simulation program as the training and verification data of the model,and the corresponding point light source coordinates were used as label data.The neural network model described in this paper is built for model parameter design and training.The simulation accuracy is compared between the simulation results obtained by the deep learning 3D coordinate positioning method based on point diffraction interference and the true value,and the feasibility,stability and high accuracy of the deep learning neural network model in this paper were verified.A deep learning 3D coordinate positioning system based on point diffraction interference was built,and the experimental accuracy was verified on a CMM.By comparing the experimental accuracy of the three-dimensional coordinate positioning method in this dissertation and the three-dimensional positioning point diffraction interference method based on polarization phase shift,it can be seen that the maximum coordinate measurement error of the method in this dissertation is 0.82 μm,and the maximum coordinate measurement error of the polarization phase shift point diffraction interference method is 3.56 μm.It is further shown that the neural network model of 3D coordinate localization presented in this paper is robust and high precision.