多视图三角化中特征点噪声尺度的自适应估算

鲁棒性多视图三角化方法通常借助重投影误差经验阈值来剔除图像对应中的错误匹配,该经验阈值的选取直接影响三维重构场景点的数量和精度。在分析图像特征点定位噪声及对极传递几何原理的基础上,建立对极传递过程不确定性的传递模型,提出一种基于核密度估计的最优噪声尺度估算方法,并将该噪声尺度作为多视图三角化中错误匹配筛选的依据。实验结果表明,该方法可以获得准确的噪声尺度估计,从而有效提升多视图三角化方法的三维重构质量。

Adaptive Estimation of Noise Scale in Feature Localization for Multi-view Triangulation

To eliminate the influence of outliers in image correspondence, robust multi-view triangulation algorithms usually rely on an empirical reprojection error threshold to identify and remove outliers. The selection of such threshold is critical to both the quantity of successfully reconstructed scene point and its accuracy. Based on an analysis of the noise in feature point localization and the geometry of epipolar transfer, this paper derives the uncertainty propagation model in epipolar transfer. A novel noise scale estimation approach based on kernel density estimation is then proposed and the estimated noise scale is further incorporated into robust state-of-the-art multi-view triangulation algorithm. The experimental results demonstrate that our proposed method is able to obtain accurate estimation of noise scale and improve the performance of multi-view triangulation algorithm significantly.