基于神经网络的锥光全息测量系统校准技术研究

Design and uncertainty analysis of non-contact measurement system based on conoscopic holography

为了解决锥光全息非接触式测量系统的输出非线性问题,提出采用三层BP神经网络方法来实现系统特性的非线性校准,充分利用神经网络自身具有的自学习和泛化能力,既避免了线性校准技术带来的原理误差,也避免了采用硬件校准方式的复杂性。大大降低了系统的测量误差,实验结果表明,校准前系统的最大测量误差为4.2μm,校准后的最大测量误差为1.5μm。

To eliminate the nonlinearity of conoscopic holography non-contact measurement system,the method of nonlinearity calibration based 3 layer BP ANN was proposed in this paper,which utilized ANN＇s ability of self-learning and generalization.It avoided the principle error of linearity calibration and complexity of calibration by hardwires.This method reduced measurement error of the system.From experiments,the maximal error before calibration is 4.2 μm and the maximal error after calibration is 1.5 μm.