【摘要】 鱼眼镜头成像立体视觉系统在微小型机器人视觉导航和近距离大视场物体识别与定位中有着广泛的应用 .尽管鱼眼镜头摄像机具有很大的视场角 (接近 180°) ,但同时也引入严重的图像变形 ,常规的摄像机标定方法无法使用 .该文提出一种标定鱼眼镜头摄像机立体视觉系统的方法 .在鱼眼镜头变形模型的基础上 ,通过考虑鱼眼镜头成像的径向变形、偏心变形和薄棱镜变形 ,建立了鱼眼镜头成像的精确成像模型 ;然后 ,利用非线性迭代算法 ,精确求解摄像机外部参数、内部参数 .实验表明 ,使用该方法得到的立体视觉系统参数满足精确恢复大场景稠密深度图的要求 .更多还原

【Abstract】 In the application of mini-micro-robot visual perception, there is a need for fish-eye lenses for capturing wide field of view for navigation. Though fish-eye lenses provide a wide field of view (～180°), they introduce significant distortion in images and the acquired images are quite warped, which makes conventional camera calibration algorithms no longer work well. This paper presents an accurate calibration framework for fish-eye lens (a high distortion lens) camera stereo vision system.The accurate calibration model is formulated with radial distortion, decentering distortion and thin prism distortion based on the fisheye deformation model. Using fish-eye and non-linear camera model, the author employs Levenberg-Marquardt method to realize precise non-linear calibration for wide-view-scene dense depth image recovery.更多还原