光子学报, 2014, 43 (5): 0512002, 网络出版: 2014-06-03
基于弱透视成像模型的目标三维姿态测量
Measurement of 3D Pose of Objects Based on Weak Perspective Model
摘要
为了准确获取空间目标跟踪、视觉导航等领域中目标的三维姿态,进行了目标三维姿态单目视觉测量方法研究.提取图像目标的典型特征点构造出直角三角形,并通过其边长比例先验信息以及弱透视成像模型推导出目标三维姿态的单目解算算法.与传统测姿方法相比,该算法在相机焦距等内参量未知条件下依然可解算姿态,增大了测姿应用范围;与传统迭代测姿方法相比,避免了循环迭代求解过程,无需设置迭代初值,提高了解算效率.数值仿真试验结果表明目标在离相机1~3 km成像时姿态测量误差低于1.5°;实际图像序列测量结果表明目标俯仰角和偏航角测量结果拟合残差小于1°,翻滚角拟合残差小于2°.实验验证了算法的正确性和稳定性,表明该算法在内参量未知条件下能有效测量中远距离成像目标三维姿态.
Abstract
In the fields of tracking target and navigation based on vision, to measure the 3D pose of rigid objects besides the special location, the research of measuring the 3D pose of the object based on mono-view was performed. An algorithm of measuring 3D pose was proposed from the weak perspective model and the information of right-angled triangle formed by characteristic points of object that was the ratio of side-length. The algorithm was applicable under the condition that the interior parameters were unknown and it avoided iteration. The result of mathematical simulating experiment shows that the error of the 3D pose angles are less than 1.5° when the distance from object to camera is in the range of 1~3 km.The result of actual images shows that the remnants difference between the raw results and smooth results of pitch and yaw are less than 1°, and of roll angle is less than 2°. The experiment proves the correctness and stability of the proposed algorithm, which shows that the 3D pose of object can be measured effectively in the proposed algorithm, which can be perfectly applied to the objects with long distance and small visual angle.
赵汝进, 刘恩海, 张文明, 赵连军. 基于弱透视成像模型的目标三维姿态测量[J]. 光子学报, 2014, 43(5): 0512002. ZHAO Ru-jin, LIU En-Hai, ZHANG Wen-Ming, ZHAO Lian-jun. Measurement of 3D Pose of Objects Based on Weak Perspective Model[J]. ACTA PHOTONICA SINICA, 2014, 43(5): 0512002.