摘要
通过在不同的方位所拍摄的两幅照片,根据双目视觉匹配原理,以循环淘汰式鲁棒性匹配结果做为种子点,用唯一性和连续性约束实现区域扩散的对应点匹配,并利用相机非线性优化方法的标定结果和SFM重建算法来恢复物体三维结构信息。本方法仅仅利用数码相机作为图像传感器来测量物体的三维模型,避免了其他的测量方法操作复杂,测量物体尺度限制性等,通过照片就能够直接生成面向反求工程的点云数据。试验结果表明本文的方法能够达到理想的精度,是有效可行的。
Two photographs produced by a digital camera at different positions realize correspondence based on region propagation by uniqueness and constraint, whose greed comes from the result of robust matching of eliminated outliers gradually and circularly. The method can recover 3D structure information by using the result of camera non-linear calibration and SFM algorithm. The digital camera as a kind of image sensor can measure three dimensions, avoid complicate operations and limits of object scale and directly produce cloud data by photographs for reverse engineering. Experiment results show that good preci- sion can be achieved and that it is effective.
出处
《机械科学与技术》
CSCD
北大核心
2005年第11期1362-1365,共4页
Mechanical Science and Technology for Aerospace Engineering
基金
国家863基金项目(2001AA421160)资助
关键词
反求工程
图像匹配
点云数据
相机标定
reverse engineering
image matching
cloud data
camera calibration
