一种不同高度目标的视觉定位新算法

本文提出一种针对平面上方高度已知目标的视觉定位算法。该方法避开了复杂的摄像机参数标定过程,直接运用射影几何的基本性质实现了平面内目标的单目视觉定位。在此基础上,进一步运用成像深度的多义性实现对高度已知目标的平面坐标定位。最后进行了初步的精度分析,给出了算法应用的建议。该定位算法无需预先知道摄像机的参数,也无需对姿态进行特别的规定,可以根据实际场景来布局定位参考点,降低定位误差。实验表明其具有实现简单,精度高的特点。

New visual localization algorithm for targets with different heights

A new visual localization algorithm is proposed for the targets whose heights above a given plane are known.Without complex camera calibration procedure,we first realize monocular localization directly for the targets on one plane according to the basic property of projective geometry.Based on the algorithm,further localization for targets with known heights is realized by utilizing the multi-meaning of the image depth.At last,the localization precision is analyzed primarily and some application suggestions are given.This algorithm does not need any priori camera parameters or special camera gesture.And the localization errors can be reduced by distributing the reference points properly according to actual environment.Experiments show that the algorithm is easy to use and has high precision.