顾及近景图像特征点误差各向异性的三维空间重构

Three dimensional space reconstruction considering the feature points' error anisotropy of close range images

本文探讨了一种考虑近景图像特征畸变的三维空间重构方法,在分析了特征点误差不确定性分布特点的基础上,在运动恢复结构的技术框架下,利用图像之间的仿射变换,将特征点误差的各向异性分布映射为各向同性,同时构造顾及特征点邻域误差各向异性的误差方程,通过迭代计算得出像片的位置和姿态及待测目标三维空间结构的最优解。最后利用多视角近距离拍摄的多幅图像进行三维重构实例验证,通过与传统方法对比分析,验证该方法的适用性和可靠性。

According to the characteristics of the feature points' perspective distortion of the close-range image,based on the principle of bundle adjustment in the structure from motion,that the synchronization calculating of the camera parameters and the three-dimensional point coordinates,a method of calculation of three dimensional structure was proposed that considering the error anisotropic of feature point.Based on the affine transformation principle between the close-range image,the anisotropic distribution of the feature point error is mapped to isotropy,the objective function was constructed considering local anisotropy,the optimal solution of the camera parameters and 3 Dpoint coordinates were obtained.Finally,the experiment of the 3 Dreconstruction with multi view close range image sequence is carried out,the validity and reliability of the proposed method are verified by comparison and analysis.