Stoilov算法中相移量的计算依赖于采集到的变形条纹图的光强,存在对光强的减法、除法和开方等运算,使相位计算时在某些位置出现分子分母为零或开方为复数等奇异现象,导致相位展开无意义或出现超大误差。从而使三维面形重构出现畸变、失真,甚至无法进行三维重构。大量实验说明这些奇异点都是孤立的点,本文提出一种基于局域均值滤波的方法对Stoilov算法进行改进,首先对奇异点进行标记,然后依次用其周围8个点中的非奇异点的平均值代替该奇异点的值。实验结果表明,该方法在考虑传感系统带来的误差的情况下,有效抑制了奇异现象,提高了三维测量精度。

In Stoilov algorithm, phase shift amount calculation depends on the intensity of distort patterns. Because subtraction, division and extraction exist in its mathematical model, it will cause some abnormal phenomena somewhere such as the zero-denominator and plural phase and so on, which leads to the incorrect phase unwrapping. The reconstructed object will be misshappen or anamorphic, or even the measured object can not be reconstructed. Many experiments show that these abnormal points are isolated.A new method based on median filtering is proposed to improve Stoilov algorithm. The abnormal points are found and are set to a constant. Then the abnormal points are replaced by the average values of eight normal points around them. Experiment results show that considering the error caused by sensing system, the abnormal phenomena can be restrained effectively and the precision of three-dimensional measurement is improved.