基于结构光的偏振主被动融合三维重建方法  

Three-dimensional Reconstruction Method Integrating Active and Passive Polarization Based on Structured Light

作　　者：李宜峰 周顺[1] 李庞跃 刘桦辰 郑欣妍 程进 孙雪平[1] 刘卫国[1] LI Yifeng;ZHOU Shun;LI Pangyue;LIU Huachen;ZHENG Xinyan;CHENG Jin;SUN Xueping;LIU Weiguo(School of Optoelectronic Engineering,Xi′an Technological University,Xi′an 710021,China)

机构地区：[1]西安工业大学光电工程学院,西安710021

出　　处：《光子学报》2024年第12期44-55,共12页Acta Photonica Sinica

基　　金：陕西省教育厅科研计划(No.21JY017)。

摘　　要：针对单一模式不足,提出了一种结构光主动与偏振被动融合的三维重建方法。首先通过加权平均去噪处理偏振信息,减少噪声的影响;之后利用结构光成像获取的深度图校正原始偏振信息中的方位角引起的歧义;最后,采用网格拉布拉斯算子优化处理表面凹陷问题,并结合结构光测得的深度实现融合算法,实现更精确的三维重建。测试结果表明,在对被测物体进行测量时,所测得的点对距离与实际测量的点对距离的绝对误差控制在0.6 mm以内,而相对误差保持在2.5%左右。本研究为无纹理表面物体的三维重建提供了一种创新的方法和思路,解决了单一模式的不足。Traditional three-dimensional imaging methods can be categorized into active imaging and passive imaging based on the presence of illumination sources.Structured light technology,as a form of active imaging,does not perform satisfactorily in imaging surfaces of objects without texture or lacking texture.Passive polarized three-dimensional imaging technology offers an effective imaging method for such objects,yet it still encounters challenges related to sensitivity to noise and ambiguity in reconstruction.In response to the limitations of a single mode,this study proposes a three-dimensional reconstruction method that combines active structured light with passive polarization.The initial step in obtaining polarised images is to rotate a polariser to capture different polarisation states.Subsequently,the images are subjected to denoising and the calculation of polarization degree,polarization angle,and zenith angle,to reduce noise's impact on polarization information significantly.This results in an ambiguous azimuth angle.To address this ambiguity,structured light is employed to obtain a point cloud and depth images of the object under examination.The gradient information of the depth image in this state is obtained and used to rectify the polarization gradient information derived from the polarization information,thus eliminating ambiguities caused by the polarization angle.Subsequently,the corrected gradient information is integrated to obtain a rough relative depth map of the object,which exhibits surface roughness at this stage.To address this,the mesh data extracted from the obtained rough relative depth map is subjected to smoothing,thereby acquiring a smooth relative depth map.Finally,the fused absolute depth map of the object is obtained by combining the original point cloud from structured light with Poisson reconstruction.In this study,key points on the measured object were randomly selected and the actual physical distances between these points were compared with the distances computed by the reconstru

关 键 词：三维重建 偏振成像 结构光 去噪 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]