三维物体视觉测量重构中解决遮挡问题的方法研究  

Research of Eliminating Occlusion in Visual Construction of Three-Dimensional Objects

作　　者：何炳蔚[1,2] 林东艺[1] 陈志鹏[1] 丁辉[1] 

机构地区：[1]福州大学机械工程及自动化学院,福建福州350108 [2]天津大学精密测试技术及仪器国家重点实验室,天津300072

出　　处：《中国激光》2011年第7期196-205,共10页Chinese Journal of Lasers

基　　金：国家自然科学基金(50605007);天津大学精密测试技术及仪器国家重点实验室开放基金;福建省高校新世纪优秀人才支持计划(XSJRC2007207)资助课题

摘　　要：利用视觉技术自动获取未知物体三维模型的过程中,如何有效解决视觉遮挡问题是研究的难点。因此,提出一种能够合理解决视觉遮挡的视觉规划新方法,该方法将物体重建过程分成360°侧面重建和上表面重建两个步骤。侧面重建中,首先预测出遮挡部分的轮廓模型,结合可视空间确定出消除遮挡的视点位置。当物体侧面信息获取完整后,通过分析物体上表面边界点的可视性,将视觉系统能够获得最大上表面信息的空间位置定义为最优翻转视点。针对上表面重建过程中的遮挡问题,提出以遮挡边界连线的法矢为依据,确定解决遮挡问题的规划方案。最后通过三维实体模型的重构实验,验证所提方法的可行性。How to solve the occlusion problem effectively is a difficult subject during the progress of acquiring unknown object＇s three-dimensional（3D） digital model.So a new view planning which can cope with occlusion correctly is proposed.The reconstruction progress is divided into two steps： 360°side surface reconstruction and upper surface reconstruction.Firstly,the contour predictable model of the occlusion is acquired and the position coping with the occlusion information is obtained with the visible region.Secondly,after the completion of side surface reconstruction,the visible criterion of the boundaries of the upper surface of the object is predicted and the best turnover angle of the vision system is defined where the vision system can acquire the maximal information of the upper surface.And then,the normal vector of the line connecting the occlusion boundary points is calculated,based on which a planning scheme to solve the occlusion is proposed.Finally,the experimental result of reconstructing the 3D real model shows that the method is feasible in practical implementation.

关 键 词：机器视觉 计算机视觉 三维重构 视觉规划 遮挡 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]