基于BP神经网络的光栅莫尔条纹细分方法  

Subdivision of grating Moiré fringes based on BP neural network

作　　者：钱绍祥[1,2] 徐德爱[1] 张飞霞[1] 张志凤[1] 

机构地区：[1]镇江高等专科学校机械工程系,江苏镇江212003 [2]江苏大学机械工程学院,江苏镇江212003

出　　处：《镇江高专学报》2013年第1期47-49,共3页Journal of Zhenjiang College

基　　金：镇江高等专科学校校级科研课题(2012070618)

摘　　要：一种基于BP神经网络的光栅莫尔条纹细分方法可以提高光栅莫尔条纹的细分精度。在1个莫尔条纹信号周期内采集多个样本点进行差值细分,利用3层BP神经网络,将获得的差值作为网络输入,该样本点在1个栅距内的微位移量作为目标输出,建立合理的网络模型,与DSP微处理器相结合实现莫尔条纹细分。通过对样本点的学习,莫尔条纹细分精度可以达到0.001微米级。同时对非样本点进行验证,实验结果表明该系统具有可行性。Subdivision of grating Moire fringes based on BP neural network can be used to enhance the accuracy of subdivision is carried out. The multiple sample points in one moir6 fringes＇ signal period were taken out and differ- ence subdivision. Using three-layer BP neural network, these difference values corresponding to the multiple sam- ple points were regarded as the input of network, and the micro displacement of the sample point in a grating pitch was taken as the target output. The rational neural network model was established and combined with DSP to do moir6 fringes＇ subdivision. Through the learning of sample point, Moire frinze subdivision Drecision can be up to 0.001 μm level. The experiment of non-sample points proved that the system was feasible, and had application value.

关 键 词：光栅 莫尔条纹 细分 BP神经网络 

分 类 号：TB183[一般工业技术]