A Batch Reconstruction Algorithm of Multi-view Images Using Image Triplets

3D reconstruction of unordered multi-view images is very sensitive to noise. Error matching relations will affect the accuracy of the reconstruction or even lead to failure. A robust batch reconstruction algorithm is proposed in this paper, first the triplets which may contain mismatches are removed using closed cycle constraint, and then the trifocal tensor constraint in triplet is used instead of the epipolar constraint of the traditional algorithm, also the linear programming algorithm with the l_∞ norm is used instead of the second order cone programming to calculate a global optimum of rotations and locations of all the images. An efficient Union Find algorithm is introduced into the reconstruction to exact the multi-view matching points, and the 3D points are computed using the iterative linear triangulation. Experimental results show that the proposed method performs satisfactorily in terms of reconstruction efficiency and accuracy.