3D Reconstruction of Line Features from Multiple Views

This paper uses line features as the basic elements for3D reconstruction of real scenes.Firstly, the average geometric reprojection error is used as the objective function,assuring the optimization of final results；secondly,an analytic correlation method of bilinear constraint for noisy 3D lines represented by Plücker coordinates is proposed,and both its theoretical proof and geometrical interpretation are provided； thirdly,a minimal representation for 3D lines with 4degrees of freedom is used during the bundle adjustment,by which the over-parameterization problem is avoided,and the algorithm convergence is enhanced and the accuracy of the final results are substantially improved.Experiments with synthetic data and real images demonstrate the feasibility and accuracy of our proposed method.