本文根据分层制造对准系统的精度要求，研究了改善精度的软件补偿措施: 设计了 SUSAN和多帧平均降噪结合的无损滤波算法以减小滤波过程对标记图像产生的畸变；以频域能量谱函数作为系统的聚焦测度函数实现自动聚焦，消减了由离焦引入的对准误差；以校正残差均方差的大小和稳定性为依据确定了三次多项式畸变模型，校正后畸变引入标记定位误差小于 0.25 μm；通过对摆动轨迹的拟合和误差补偿使系统机械不稳定性造成的对准精度损失小于 1.3 μm。通过 Delphi和 Matlab混合编程完成了上述功能的软件实现和集成。

The software compensation techniques are investigated to improve the alignment accuracy in the alignment system for layered manufacturing process. The filtering algorithm with little degradation on image, which combines the SUSAN with multi-frame averaging, is introduced to alleviate the mark image distortion due to the filtering processing. The focus measure function based on Power Spectrum in Frequency Domain (PSFD) is adopted as the auto-focus function of the system and then the positioning error due to defocus is reduced or eliminated. Concerning the calibration residual error and its repeatability of various models, the cubic polynomial model is used to describe and calibrate the geometric distortion of mark image and the alignment error due to image geometric distortion is less than 0.25 μm after calibration. The alignment error due to mechanical instability of the alignment setup is reduced to less than 1.3 μm through mechanical swing trace fitting and error compensation. Finally, adopting the hybrid programming method with Delphi and Matlab, the software functions mentioned above are implemented and integrated.