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%。 For the scanning probe microscope scanning image display size and the actual size there is a big error, based on image processing and artificial neural network calibration method. The feature matrix of standard raster image is extracted through image processing, and the feature matrix and the real matrix are trained by artificial neural network to predict the voltage value actually applied to the x and y axes of the scanned image. Applied to a certain AFM calibration software, the x_y surface distortion of the scanned image was reduced from 7.2% to 2.44%.