基于外点检测与校正的填充射影分解算法

针对三维重构中存在的数据缺失和遮挡问题，提出可处理缺失数据的填充射影分解算法，利用子空间约束与对极几何约束进行矩阵拟合并填充缺失数据，通过奇异值分解得到射影运动与结构参数。为克服该算法对噪声和外点的敏感性，结合RANSAC算法和三角形法对其进行外点检测与校正。实验结果表明，加入外点校正后的算法可提高射影重构的鲁棒性，降低误差，具有较高的实用价值。

Filling Projective Decomposition Algorithm Based on Exterior Point Detection and Correction

Aiming at the problems of data missing and occlusions in 3D reconstruction, this paper proposes a filling projective decomposition algorithm which can handle missing data. Sub-space and epipolar constraints are used to fit the measurement matrix and fill the missing data. The projective motion and structure are recovered by Singular Value Decomposition(SVD). To solve the problem that the method is sensitive to noise and exterior point, RANSAC algorithm and triangulation algorithm are employed to detect and correct the exterior point. Experimental results indicate that the algorithm can strengthen the robustness and reduce error for projective reconstruction through correcting the exterior point, and it has great application value.