一种逐周期条纹背景自适应去除算法

A Periodic Stripe Background Adaptive Removal Algorithm

摘要:

利用傅里叶轮廓术(FTP)可实现对单帧光栅投影条纹图像的相位提取，恢复被测物体三维信息。该方法具有实时性高的特点，但在频域滤波时易受到不均匀背景分量的干扰。现有投影条纹图背景去除方法在遇到条纹图背景分量不均匀时，存在去除效果差，严重影响重建精度的问题。针对该问题，通过实验发现单个条纹周期内的背景分量变化很小，可近似作为常量处理；结合两次希尔伯特变换能有效去除信号直流分量且保持相位信息稳定的特性，提出一种自适应的分周期条纹背景滤除方法。该方法首先对光栅投影图像中的条纹峰值进行提取，分割出逐个条纹周期；再对逐条纹周期分别进行两次希尔伯特变换，自动去除该周期内的背景分量，实现条纹背景分量的逐周期自适应去除。该方法能有效克服条纹投影图像背景分量不均匀的问题。实验结果表明，该方法相对于傅里叶轮廓术可更好的消除条纹背景分量影响，对复杂表面物体的重建异常点少，精度优于0.1mm。

Abstract:

Abstract：Fourier Profilometry (FTP) can be used to realize phase extraction of single frame grating projection fringe image and restore the 3D information of the measured object. This method has the characteristics of high real-time performance, but it is easy to be interfered by non-uniform background components in frequency domain filtering When the background component of the fringe image is not uniform, the existing methods for background removal of the projected fringe image have poor removal effect, which seriously affects the reconstruction accuracy. Aiming at this problem, it is found through experiments that the background component changes very little within a single fringe period, which can be approximately treated as a constant. Considering that the two Hilbert transforms can effectively remove the DC component of the signal and keep the phase information stable, an adaptive fractional periodic fringe background filtering method is proposed. Firstly, the fringe peak is extracted from the raster projection image, and the fringe period is segmenting. Then, two Hilbert transforms are carried out on each fringe period to automatically remove the background component in this period, so as to realize the periodic adaptive removal of the fringe background component. This method can effectively overcome the problem of the non-uniform background component of the fringe projection image Experimental results show that the proposed method can eliminate the fringe background component better than Fourier contour technique, and the reconstruction accuracy is better than 0.1mm with fewer abnormal points on complex surface objects