Research On Nonlinear Vector Tomograph Reconstruction Algorithm Based On DNN

With the continuous development of Internet of Things technology and its information industry,automatic identification of information,as the core application of the Internet of Things,has attracted more and more attention in recent years.Metal tag system is a key technology to realize automatic identification of magnetic field information of metal materials,which fills the gap in the field of metal material identification in the existing Internet of Things sensing technology.Metal tag system obtains the shape and position information of metal mark by measuring the magnetic field on metal surface.Traditional magnetic field measurement such as fluxgate and Hall sensor has the problems of low sensitivity,high cost and unable to obtain complete magnetic field information.In order to remedy the above shortcomings,this paper combines the amorphous wire sensing materials with magnetic field resolution and vector tomography technology to study the reconstruction method of vector field tomography based on the non-linear characteristics of amorphous wire.Vector tomography reconstruction technology is studied in this paper.Using projection data and related boundary information,the vector distribution information can be restored,and the magnitude and direction of magnetic field can be reconstructed.At the same time,as a metal alloy material,amorphous wires have giant magneto-impedance characteristics.By measuring the impedance of amorphous wires,the projection information of magnetic field components can be obtained,which improves the sensitivity and resolution of magnetic field detection.Aiming at the giant magneto-impedance characteristics of amorphous wire sensing materials,a non-linear transformation model between the impedance of amorphous wire and vector magnetic field was established.The sensing model and giant magneto-impedance effect of amorphous wires are analyzed.The relationship between the impedance of amorphous wires and magnetic field amplitude,magnetic field direction and alternating current frequency is simulated and analyzed by MATLAB.Three kinds of non-ferrous wires are constructed by fitting curves under different parameters.The nonlinear transformation model of crystal filament.In order to solve the nonlinearity problem of amorphous wire sensing,a reconstruction algorithm of non-linear vector tomography based on DNN network is proposed.The DNN(Deep Neural Network)model based on the projection of amorphous wire is constructed,and the mapping relationship between the non-linear projection data and the linear projection data is established to obtain the linear projection matrix of the original vector.The original magnetic field is reconstructed based on the vector tomography,and the vector image is reconstructed by MATLAB simulation.The results show that the error of reconstructing vector field by analytic method with DNN network is smaller than that of traditional shallow BP(Back Propagation)algorithm and direct analytic method.