一种基于空间编码结构光的稠密三维重建算法  

Dense three-dimensional reconstruction algorithm based on spatially encoded structured light

作　　者：陈荣 许宏丽[1,2] 杨东学 黄华[1,3] CHEN Rong;XU Hongli;YANG Dongxue;HUANG Hua(School of Computer and Information Technology,Beijing Jiaotong University,Beijing 100044,China;Beijing Key Laboratory of Traffic Data Analysis and Mining,Beijing Jiaotong University,Beijing 100044,China;Key Laboratory of Beijing for Railway Engineering,Beijing Jiaotong University,Beijing 100044,China)

机构地区：[1]北京交通大学计算机与信息技术学院,北京100044 [2]北京交通大学交通数据分析与挖掘北京市重点实验室,北京100044 [3]北京交通大学轨道工程北京市重点实验室,北京100044

出　　处：《西安电子科技大学学报》2021年第6期123-130,共8页Journal of Xidian University

基　　金：国家自然科学基金(51827813,61472029);科技部国家重点研发计划(2017YFB1201104);北京市教育委员会科技重大项目(KJZD20191000402)。

摘　　要：基于结构光的三维重建技术具有准确度高、速度快、鲁棒性好等优点,被广泛应用于生物医疗、文物修复等领域,具有重要的理论意义和研究价值。随着科技迅速发展,高速运动目标的非接触检测和测量的需求逐渐增加,对特征少、速度快的目标三维重建成为解决问题的关键。相比于时间编码,空间编码仅需拍摄单幅编码图案,更适用于运动目标的三维重建。因此,基于空间编码结构光提出了一种稠密三维重建方法。首先,针对彩色条纹结构光受物体颜色影响的问题,设计一种单色正弦条纹加伪随机点的结构光编码模式,通过伪随机点识别条纹,降低目标物体的颜色干扰;其次,针对传统条纹方法得到的点云空间分辨率低的问题,提出一种基于局部相位匹配的稠密三维重建方法,通过获取相邻条纹间的像素相位信息进行像素匹配,提高点云空间分辨率。同时,该方法仅需要一个图像对来进行解码,且不需要对投影仪进行标定和颜色校正,可提高三维重建速度。实验结果表明了该算法的可行性和有效性。The 3D reconstruction technology based on structured light has the advantages of high accuracy,fast speed,and good robustness.It is widely used in the fields of biomedicine and cultural relics restoration.It is of important theoretical significance and research value.With the rapid development of science and technology,the demand for non-contact detection and measurement of high-speed moving targets has gradually increased.The 3D reconstruction of targets with few features and fast speed has become the key to solving the problem.Compared with time coding,spatial coding only needs to shoot a single pattern,which is more suitable for 3D reconstruction of moving targets.This paper proposes a novel dense 3D reconstruction method based on spatially encoded structured light.First,a structured light encoding mode composed of sine monochrome stripes and pseudo-random points is constructed.The stripes are identified by pseudo-random points to reduce the color interference;second,a dense 3D reconstruction method based on local phase matching is proposed.The spatial resolution of point cloud is improved by obtaining the phase information on pixels between adjacent stripes for pixel matching.Finally,the method can enhance the speed of 3D reconstruction,which uses only one image pair to decode,and does not need to calibrate the projector and color correction.Experimental results show that the algorithm is feasible and effective.

关 键 词：表面重建 结构光 投影仪-相机系统 傅里叶变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]