摘要
鱼眼镜头是一种具有广视场角的摄像机镜头,广泛应用于大场景视频监控领域,但其成像存在严重的非线性径向畸变,违反人类的视觉习惯,且为测量和模式识别等应用带来了不便。为了实现鱼眼镜头图像的校正,提出了一种基于几何成像模型的鱼眼镜头校正算法和技术,通过研究鱼眼镜头成像模型,根据鱼眼镜头类型采用了相应的几何模型校正图像;以"等距投影"模型为例提出了校正公式和实现方法。最后针对工程应用的实时性要求,提出了基于缓存的快速优化方法以及在保证一定精度的前提下快速实现双线性插值的优化方法。研究结果表明,这种新方法可以高效地完成鱼眼镜头图像校正,比传统方法适用面更广,更易实现。
Fish-eye lenses with wide field of view（FOV） are widely used in the field of large scene video surveillance,but nonlinear radial distortion of fish-eye image violated the human visual habits and brought inconvenience to measurement and pattern recognition applications.In order to correct the fish-eye images,a new correction method and technique based on geometrical imaging model that corrects the images form different types of fish-eye lens by corresponding geometrical imaging model was proposed.Also,the implementation method of equidistant imaging model was developed.Finally,the optimization methods based on the cache and a fast bilinear interpolation implementation ensuring the accuracy aiming at the real-time application were presented.The results indicate that the new method works efficiently in a larger scale,and easier to implement.
出处
《机电工程》
CAS
2013年第10期1268-1272,共5页
Journal of Mechanical & Electrical Engineering
关键词
鱼眼镜头
径向畸变校正
图像处理
视频监控
fish-eye lens
radial distortion correction
imaeg process
video surveillance
