基于一维经验模态分解的非等步在线三维面形测量  

An unequal step on-line 3D shape measurement based on 1D empirical mode decomposition

作　　者：郑素珍[1,2] 曹益平[1] 

机构地区：[1]四川大学光电科学技术系,四川成都610064 [2]西南石油大学理学院,四川成都610500

出　　处：《光电子．激光》2014年第8期1527-1534,共8页Journal of Optoelectronics·Laser

基　　金：国家"863"高技术研究发展计划(2007AA01Z333);国家重大专项(2009ZX02204-008);四川省学术和技术带头人培养基金(07GRC-01);四川省教育厅科技项目(13ZA0191)资助项目

摘　　要：提出一种基于一维经验模态分解(EMD,empirical mode decomposition)的像素匹配方法,应用于在线三维面形测量。首先由计算机产生一固定正弦光栅条纹图,利用DLP投影仪投射到被测物体表面,物体在线运动将产生等效相移,由CCD相机采集5帧任意步长的受物体面形调制的变形条纹图。由于物体的运动使得几帧变形条纹图中的物体像素不对应,因此需要对变形条纹图进行像素匹配。通过对采集到的5帧变形条纹图进行一维EMD,可以得到与待测物体形貌变化一致的模态图像分布,然后通过大津法对其进行二值化处理,并截取可靠性较高的部分作为像素匹配模板进行像素匹配,计算出相应的等效相移量。最后利用五步非等步相移算法解相,并恢复出被测物体的三维形貌。计算机仿真与实验验证了本文方法的有效性。A new pixel matching method is proposed based on one-Dimensional empirical mode decompo- sition（1D-EMD） and applied to the on-line 3D shape measurement. Firstly, a fixed sinusoidal grating fringe pattern generated by computer is projected onto the measured object surface by a digital light pro- cessing （DLP） projector. The object on pipe-line movement will produce equivalent phase shifting. Then a CCD camera with any step captures five frames of fringe patterns modulated by the object^s surface shape. Due to the movement of the measured object,this makes the points in these deformed fringes not be one-to-one corresponding. The pixel matching must be done. In this paper, the five deformed patterns are decomposed with 1D-EMD method, and we can obtain the mode distribution which is consistent with the objectp surface shape. And then Otsu algorithm is used to make the binarizatiom The area with higher reliability is segmented to be regarded as a template to realize the pixel matching. And the phase shifting algorithm with fine unequal steps is used to reconstruct the surface information of the measured ＂object. The simulations and experiments prove the feasibility of this method.

关 键 词：测量 非等步相移 在线测量 经验模态分解(EMD) 像素匹配 

分 类 号：O438.2[机械工程—光学工程]