Grating subdivision method based on radial basis function neural network

In order to develop the grating displacement measuring system with higher subdivision accuracy and displacement tracking speed, a grating subdivision method based on radial basis function neural network was proposed. The multiple sample points in one moiré signal period were taken out using three-layer RBF neural network. The tangent values corresponding to the multiple sample points were taken as the input of the network and the micro displacement of the sample point in a grating pitch was regarded as the target output. The rational neural network model was established and combined with DSP to achieve the moiré fringe subdivision. Through the fractional learning of sample point, it is demonstrated that the high precision subdivision can be realized only with a few neurons. The structure of this neural network is simple and the ability of nonlinear approximation is powerful. The experiments of non-sample points prove that the system is feasible, and has application value.