作者：赵鹏亮^1,2, 梁晋^1,2, 徐劲澜^1,2, 武国庆³, 孟繁昌^1,2

作者单位：1. 西安交通大学机械工程学院 机械制造系统工程国家重点实验室, 陕西 西安, 710049;
2. 广东顺德西安交通大学研究院, 广东 佛山 528300;
3. 中航工业洛阳电光设备研究所, 河南 洛阳 471009

关键词：面部测量;激光散斑;数字图像相关法;图像去边缘化

摘要：

针对医学等相关领域对面部轮廓测量的需求，提出一种主动激光散斑投射的面部点云获取方法。该方法结合数字图像相关法和双目立体视觉原理，能够快速重建面部三维轮廓。首先，介绍数字图像相关法的基本原理；其次，提出一种匹配前的散斑图像预处理方法，去除散斑图像的边缘化区域，提高散斑匹配质量；最后，计算面部三维点云。为验证该方法的可行性及最终重建效果，设计面部扫描重建及误差实验，结果表明：该方法具有体积小和不受光强影响的特点，能够较好地重建出面部模型，整体测量误差为0.2 mm。

Research on active laser speckle projection for facial measurement

ZHAO Pengliang^1,2, LIANG Jin^1,2, XU Jinlan^1,2, WU Guoqing³, MENG Fanchang^1,2

1. State Key Laboratory for Manufacturing Systems Engineering, School of Mechanical Engineering, Xi'an Jiaotong University, Xi'an 710049, China;
2. Guangdong Shunde Xi'an Jiaotong University Academy, Foshan 528300, China;
3. Institute of Electro-Optical Equipment, AVIC, Luoyang 471009, China

Abstract: In order to solve the problem of facial measurement in medical and other related fields, a method to acquire the facial point cloud based on active laser speckle projection is proposed. This method can rapidly reconstruct the three-dimensional facial contour by combining stereo vision with digital image correlation method. Firstly, the fundamental of digital image correlation method is introduced, and then a preprocessing method of speckle image before matching is proposed to remove the marginalized area of the speckle image for high quality speckle matching. Finally, according to the speckle matching, the facial three-dimensional point cloud is calculated. To verify the feasibility and the final reconstruction effect of the method, the facial scan reconstruction and error experiment are designed and conducted. Experimental results demonstrate that the method has small measuring device size and light robustness. It can reconstruct the facial model well, and the overall measurement error is ±0.2 mm.

Keywords: facial measurement;laser speckle;digital image correlation method;image marginalization removal

2017, 43(9): 8-12  收稿日期: 2016-12-23;收到修改稿日期: 2017-02-10

基金项目: 国家自然科学基金项目（51421004）；广东省公益研究与能力建设专项资金（2014A010104003）

作者简介: 赵鹏亮(1991-),男,河南许昌市人,硕士研究生,专业方向为三维光学测量。

参考文献

[1] 钟世镇. 数字化虚拟人体研究现状和展望[J]. 解放军医学杂志,2003,28(5):385-388.
[2] 潘红艳,柳杨华,徐光祐. 人脸动画方法综述[J]. 计算机应用研究,2008,25(2):327-331.
[3] 熊耀阳,陈晓波,孙健,等. 结构光投影面部三维测量系统的研制[J]. 上海交通大学学报(医学版),2009,29(7):837-841.
[4] 杨宇,阚凌雁,于佳,等. 基于激光扫描的人脸三维重建方法[J]. 红外与激光工程,2014,43(12):3946-3950.
[5] 鹿乐,周大可,胡阳明. 基于特征分块的三维人脸重建和识别[J]. 计算机应用,2012,32(11):3189-3192.
[6] 葛宝臻,赵丹,田庆国,等. 三光带激光三维人脸扫描方法[J]. 2010,37(6):1594-1598.
[7] 唐正宗,梁晋,肖振中,等. 用于三维变形测量的数字图像相关系统[J]. 光学精密工程,2010,18(10):2244-2253.
[8] SAVIC V, HECTOR L G J, FEKETE J R. Digital image correlation study of plastic deformation and fracture in fully martensitic steels[J]. Experimental Mechanics,2010,50(1):99-110.
[9] KARTAL M E, MULVIHILL D M, NOWELL D, et al. Determination of the frictional properties of titanium and nickel alloys using the digital image correlation method[J]. Experimental Mechanics,2011,51(3):359-371.
[10] 邓祥. 多分辨率与视差梯度相结合的快速区域匹配[J]. 山西电子技术,2014(6):9-11.
[11] MCDONNELL M J. Box-filtering techniques[J]. Computer Graphics and Image Processing,1981,17(1):65-70.