基于神经网络的反求测量系统的标定

Calibration Methods Based on Neural Network for Cross-sectional Imaging Measurement System

在层去图象法测量系统中，由于诸多因素的影响，物体的空间坐标与截面图象坐标之间存在着复杂的非线性映射关系。如果采用完全理想条件和线性几何失真方法来标定系统，则会影响测量精度，为此提出了一种基于神经网络的标定方法，显著地提高了测量系统的精度。

There is complex nonlinear mapping between the object space and image space due to many factors in cross-sectional imaging measurement system. The measurement accuracy will be low if the ideal or linear conditions are used to calibrate the system. A calibration method based on neural network is presented. The results show a remarkable improvement of the measurement accuracy.