基于时空傅里叶变换的动态过程三维面形测量研究

With the merit of only one single frame of deformed fringe pattern being needed to reconstruct three-dimensional (3D) shape of a tested object, 3D shape measurement technology based on structured light projection and Fourier transform fringe analysis is the first choice of the reconstruction method for a dynamic process. But, when the 1D or 2D Fourier transform is usually carried in spatial domain, the employed weighted band-pass filter is difficult to perform automatically, and furthermore its limited band width will ruin or inhibit the measured high frequency information, which eventually makes the tested dynamic object of this technology can only be flat, smooth, continuous on its intrinsic surface distribution. In order to overcome this problem and extend the application filed of this kind technology, this proposed method in this project will project a multi-period phase-shifting fringe onto the complex and dynamic tested object. After a series of 3D (XYT) deformed fringe patterns being recorded, Fourier transformation and filtering process will be carried in the temporal domain (along the time axis T on each pixel) to calculate the corresponding 3D wrapped phase distribution of the tested dynamic object, meanwhile, the traditional 2D Fourier transform fringe analysis on the spatial domain (XY) is executed on ONLY one frame of deformed fringe pattern to retrieve the correct phase order of the wrapped phase and to guarantee the whole phase unwrapping process will be successfully completed is 3D space, and then the changing 3D shape of the measured object will be restored. The proposed method analyzes the fringe pattern in temporal domain, which will not damage the object’s spatial spectrum, so it can effectively keep the corresponding high frequency information. This method can achieve 3D shape measurement of a complex and dynamic objects with large gradient, spatial discontinuity and even some isolated area. So, it can be applied to dynamic 3D shape measurement and deformation analysis for moving and changing assembled mechanical parts in the intelligent industrial manufacturing process.

因为只需要单帧变形条纹图来重建被测物体的三维面形，基于结构光投影和傅里叶条纹分析的三维测量技术成为动态过程三维重建的首选方法。但由于所采用的傅里叶变换通常在二维空域内进行，带通的加权滤波器难于实现自动化，并且其有限的带通宽度会抑制、丢失待测的高频信息，最终使得应用对象只能是本征面形分布平缓、平滑连续的动态物体。为了克服该问题，拓展该技术的应用，本项目将研究采用多周期相移的时空傅里叶条纹分析新方法，在获得待测动态物体表面一系列时间空间相移的变形条纹图像集后，分别在时域和空域中做傅里叶条纹分析操作，获得对应的三维截断相位分布，进而完成动态三维测量。该方法在时域进行傅里叶变换分析，将不损伤物体空间频谱，能有效保留其对应的高频信息部分；能实现陡度大、空间不连续甚至有孤立区域的复杂动态物体三维面形测量，有望在诸如智能工业制造中组合机械部件运动形变等复杂动态物体的三维测量及形变分析中得到推广应用。

为改进和提升基于结构光投影和条纹分析的动态三维面形测量现有技术，实现复杂动态物体三维表面形状连续变化过程的准确记录和重建，项目组申请并获准了该项目的研究。. 经过项目组成员的分工合作，项目顺利达到了预期的研究目标：研究提出了基于结构光投影和时空傅里叶条纹分析的复杂动态三维测量新方法(TSFTP)完整的技术方案和测量系统，研制了对应的2套原理装置样机，可实现大陡度、空间不连续甚至有孤立区域的复杂动态过程三维面形测量；研究提出了基于格雷码图案投影的高速动态三维面形重建方法，可实现高噪声干扰的复杂飞散动态场景高速(542fps)、高效、高鲁棒三维形貌重建；研究提出了基于相移条纹投影的实时动态三维面形重建方法，成功实现了动态场景60fps的实时三维重建显示；研究提出了一些在基于结构光投影的三维面形测量技术中具有明显的使用价值的单元技术；完成了以冲撞破裂、机械部件往复运动、积木坍塌等复杂动态场景为典型案例的应用研究。. 项目成果目前公开发表了期刊论文25篇(SCI收录11篇，EI收录19篇)，会议文章4篇；申请技术发明专利14项，其中8项获得授权。项目组成员受邀在国际国内会议上作了11次邀请报告，6次分会场口头报告；受国内大学或科研院所邀请作了5次学术讲座；在2019年(第301场)中国工程科技论坛上作了邀请报告。项目成果4次受邀在博览会上路演展示和样机推介。受项目资助，10名硕士生已经完成学位论文研究，获得硕士学位；1名硕士、4名博士研究生的学位论文研究进展顺利。. 项目研究提出的新方法原理及应用研究具有明显的科学意义和应用价值，能迎合制造业变革中复杂面形动态形变检测的需求。研究成果将有望在关乎国家重大战略和国计民生的军民领域专用装备形貌和形变测试、智能制造领域和VR/AR领域中的复杂动态场景三维形貌重建及形变分析中推广应用。