Feng Shijie, Chen Qian, Zuo Chao, Tao Tianyang, Hu Yan, Asundi Anand
Opt Express. 2017 Jan 23;25(2):540-559. doi: 10.1364/OE.25.000540.
Fringe projection is an extensively used technique for high speed three-dimensional (3-D) measurements of dynamic objects. To precisely retrieve a moving object at pixel level, researchers prefer to project a sequence of fringe images onto its surface. However, the motion often leads to artifacts in reconstructions due to the sequential recording of the set of patterns. In order to reduce the adverse impact of the movement, we present a novel high speed 3-D scanning technique combining the fringe projection and stereo. Firstly, promising measuring speed is achieved by modifying the traditional aperiodic sinusoidal patterns so that the fringe images can be cast at kilohertz with the widely used defocusing strategy. Next, a temporal intensity tracing algorithm is developed to further alleviate the influence of motion by accurately tracing the ideal intensity for stereo matching. Then, a combined cost measure is suggested to robustly estimate the cost for each pixel and lastly a three-step framework of refinement follows not only to eliminate outliers caused by the motion but also to obtain sub-pixel disparity results for 3-D reconstructions. In comparison with the traditional method where the effect of motion is not considered, experimental results show that the reconstruction accuracy for dynamic objects can be improved by an order of magnitude with the proposed method.
条纹投影是一种广泛应用于动态物体高速三维(3-D)测量的技术。为了在像素级别精确地重建运动物体,研究人员倾向于将一系列条纹图像投影到物体表面。然而,由于图案集的顺序记录,运动往往会导致重建中出现伪影。为了减少运动的不利影响,我们提出了一种结合条纹投影和立体视觉的新型高速3-D扫描技术。首先,通过修改传统的非周期正弦图案实现了可观的测量速度,这样可以使用广泛应用的散焦策略以千赫兹的频率投射条纹图像。接下来,开发了一种时间强度跟踪算法,通过精确跟踪用于立体匹配的理想强度来进一步减轻运动的影响。然后,提出了一种组合成本度量来稳健地估计每个像素的成本,最后是一个三步细化框架,不仅可以消除由运动引起的异常值,还可以获得用于3-D重建的亚像素视差结果。与不考虑运动影响的传统方法相比,实验结果表明,所提出的方法可以将动态物体的重建精度提高一个数量级。