基于人工神经网络的SPM自动标定

Automatic calibration of SPM scan image by artificial neural network

针对扫描探针显微镜中扫描图像显示的大小与实际大小存在较大误差,提出了基于图像处理和人工神经网络标定的方法。经过图像处理提取标准光栅图像的特征矩阵,用人工神经网络训练特征矩阵和真实矩阵,用于预测扫描图像实际施加在x,y轴的电压值。应用于某原子力显微镜的标定软件中,使扫描图像的x_y表面失真度从7.2%降至2.44%。

Aiming at big discrepancy between SPM scan image display size and it＇ s actual size, an automatic calibration method based on digital image processing and artificial neural network was proposed. The characteristic matrix of standard grating was obtained by digital image processing, characteristic matrix and actual matrix were trained by artificial neural network to predict voltages which applied to x and y axis of the scanner. The experiment of calibration of an atom force microscope （AFM） shows that x-y plane distortion degree decrease from 7.2 % to 2.44 %.