基于视觉标记的增强现实系统建模及配准误差问题研究  

Research on Augmented Reality System Modeling and Registration Error Based on Simple Visual Marker

作　　者：张国亮[1] 吴琰翔 王展妮[1] 王田[1] 

机构地区：[1]华侨大学计算机科学与技术学院,厦门361021

出　　处：《计算机科学》2015年第6期299-302,F0003,共5页Computer Science

基　　金：国家自然科学基金(61202468);福建省自然科学基金(2013J05091);华侨大学高层次人才科研启动项目(11BS107)资助

摘　　要：针对当前基于视觉的增强现实系统标记大、成本高的缺陷,提出一套结合简单组合标记的增强现实系统的实现方案。基于弱透视模型,以色彩和形状作为融合特征,研究了利用冗余特征实现标记物体的3D位姿还原方法。为使配准方案能够适应普通的视觉采集设备,提出根据摄像机标定时得到的扭曲参数对渲染后的模型进行再扭曲的配准修正方法。结合OpenCV和OpenGL中的空间变换的概念,给出了虚拟和物理世界配准融合的解决方案,并基于统计学数学模型,提出降低配准时的误差以及提升配准稳定性的方案。实验结果表明,提出的算法能有效地将静态标签的抖动降低至亚像素级。An image registration strategy was studied based on simple visual marker, which is furthered to build aug- mented reality system. Firstly, color and shape were selected as integrated feature to ensure the reliability of marker de- tection. Redundant features were utilized to calculate pose of visual markers based on weak perspective model. Secondly, to make image registration be suitable for common vision capture device, rendered model was redistorted according to calibration distortion parameters. Finally,based on space transfurmation in OpenCV and OpenGL,virtual and real regis- tration was discussed and implemented with visual marker detection result.

关 键 词：增强现实 位姿估计 弱透视模型 图像配准 标记检测 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]