基于鱼眼镜头的全方位视觉参数标定与畸变矫正  

Omni-Directional Vision Parameter Calibration and Rectification Based on Fish-Eye Lens

作　　者：冯为嘉[1] 张宝峰[2] 曹作良[2] 

机构地区：[1]天津大学精密测试技术及仪器国家重点实验室,天津300072 [2]天津理工大学天津市复杂工业系统控制理论及应用重点实验室,天津300384

出　　处：《天津大学学报》2011年第5期417-424,共8页Journal of Tianjin University(Science and Technology)

基　　金：国家高技术研究发展计划(863计划)资助项目(2007AA04Z229);国家科技部中芬国际合作资助项目(2006DFA12410)

摘　　要：针对利用鱼眼镜头构建的全方位视觉系统研究内外部参数标定及图像畸变矫正方法,建立成像系统模型,提出成像系统中需要标定的内、外部参数;采用改进的线性标定法标定图像中心,研究径向畸变系数及其他参数的标定方法;在参数标定的基础上,分别利用等距投影和支持向量机训练方法,对图像中的像素点及整幅鱼眼图像进行畸变矫正.实验证明,研究的标定方法可准确地标定出视觉系统的内外部参数;基于等距投影的像素点矫正可应用于精确定位视觉系统的空间位置;基于支持向量机训练方法的全图像畸变矫正可获得理想的鱼眼图像矫正效果,所提出的参数标定和畸变矫正方法将有利于与鱼眼镜头在机器视觉领域的应用.A parameter calibration and rectification method for omnidirectional vision system based on fish-eye lens was investigated in this paper.An imaging system model was established,which indicated the interior and exterior parameters.The optic center was calibrated by an improved linearization method,and the other parameters such as the radial distortion parameters kx,ky are analyzed too.On this basis,the important point and fish-eye lens image will be rectified by the equidistance projection theorem and support vector machine training,respectively.The experiments prove that the interior and exterior parameters can be calibrated accurately utilizing the calibration methods proposed.The important point rectification method based on equidistance projection theorem can be used to calculate the precise localization for vision system.An ideal rectification image can be achieved by using support vector machine training.This research will be beneficial to the extension of fish-eye lens in the field of machine vision.

关 键 词：全方位视觉 鱼眼镜头 参数标定 畸变矫正 支持向量机 

分 类 号：TN247[电子电信—物理电子学]