摘要
本文在分析计算机视觉系统中摄象机几何模型的基础上,修正了光学镜头畸变模型使之与数字成象系统相配合,提出了一个非常适合于3—D计算机视觉系统的通用高精度摄象机定标方法:增强型直接线性变换—误差拟合算法(DLTEAⅡ).实验表明,使用普通CCD摄象机的立体视觉系统经DLTEAⅡ定标后,在300mm×200mm视场和0.8m~0.9m深度范围内其三维定位的绝对精度分别为dx<0.03mm,dy<0.04mm和dz<0.09mm,相对精度为视场大小和深度的1/10000.本文所提出的定标方法在三维计算机视觉、智能机器人、工业自动化生产、近景摄影测量和军事等领域有着广泛的应用前景.
The paper presents a new high accurate camera calibration approach for 3-D computer vision system-named as DLTEAI (Direct Linear Transformation-Error Approximation I). To overcome shortcomings of digital image acquisition, we introduce a modified lens distortion model and a quadratic polynomials of system error approximation based on the posterior compensation in the DLTEAI approach. The stereo vision system with off-the-shelf CCD or TT cameras calibrated by DLTEAI has an average relative accuracy of 1/10000 over the field of view (300mm by 200mm) and the depth (0.8m-0.9m). The absolute location errors are less than 0.03mm, 0.04mm and 0.09mm in x, y and z directions respectively.
出处
《电子学报》
EI
CAS
CSCD
北大核心
1992年第11期12-18,共7页
Acta Electronica Sinica
基金
国家自然科学基金资助重大项目(NSF 9690015-01)
关键词
摄象机
定标
计算机视觉
Camera calibration, Stereo vision, Acquisition 3-D informatiion, DLT Transform, Computer vision