Abstract: The subdivision precision of MoiréFringe is restricted by its quality. In this paper, the adaptive filtering algorithm based on the neural network is used to restrain noise of MoiréFringe.The nonlinear mapping fuction is achieved by using the neural network layer. The step size of the algorithm can be adjusted dynamicly according to the signal’s frequency to meet the filtering request of the signal with a diversified frequency and make the algorithm self-adaptive.On subdividing filtered circle grating MoiréFringe signals by means of the tangent method by 512 times, it is shown that the distinguishment is 0.618″, that the maximal cumulate error is 1.236″, and that the signal quality and the subdivision precision are greatly improved.Because of its wide bandwidth and restraint of linear and nonlinear noises, the algorithm is better than general filtering way and can satisfy the preparation for the subdivision of MoiréFringe.

Key words: Moiréfringe, step size, neural network, filtering