Image-based 3D reconstruction is one of the fundamental goals of.computer vision and has many applications such as measurement,object recognition,robot navigation, virtual reality and computer cartoon.The thesis is focused on 3D reconstruction and camera calibration,the main work is summarized as follows: 1.A robust Euclidean reconstruction method based on four point correspondences is proposed.At first, affine reconstruction is obtained linearl3r from four point correspondences (The corresponding space points are not coplanar).Then Euclidean geometry of a scene can be reconstructed linearly if the camera undergoes two sets of translations,of which the c85orresponding rotation axes of camera are not parallel.And a RANSAC (RANdom SAmple Consensus)based algorithm is introduced. which is shown by experiments to be sufficiently robust. 2.A technique for the metric reconstruction based on one triple of corresponding vanishing line in three images is proposed.Firstly,the in infinite homography is computed by using the vanishing line correspondence and the modulus constraint, then the camera intrinsic parameters are determined linearly by using the property that the infinite homography preserves the image of the absolute conic.and a metric reconstruction is obtained via the calibrated camera parameters. 3.A theory and algorithms of camera calibration based on two space parallel segments are proposed.The main results are:(1)If the length ratio o f the two parallel segments is known,then from its projections across n images,(a)2(n-1) quadratic constraint equations on camera internal parameters can be obtained;(b)(5n-6) constraint equations Can be obtained if additionally the epipoles across n images are known,among which 3(n-1)ones are quadratic,(n-2)ones of third degree,and(n-1) ones of fourth degree.(2)If the length ratio of the two parallel segments is unknown, (2n-3)constraint equations are derived from n images,among which(n2-2)ones are of fourth degree,and (n-1)ones of sixth degree.In addition,the independence of the resulting constraints is verified by simulated data.Based on the above theoretical results,two practical algorithms of camera calibration are also proposed. 4.A algorithm for estimating the aircraft approach angle frl3m image sequence is proposed.Firstly, camera intrinsic parameters are computed from the image sequence of a planar template,and the epipole and the vanishing line are extracted after that. Then the aircraft approach angle is derived from these parameters.
修改评论