Research On The Key Algorithm Of 3-D Reconstruction Technology For Panoramic Stereo Spherical Vision System

Panoramic Stereo Spherical Vision System(PSSVS)is a new visual design which can acquiring panoramic image while get three-dimensional(3D)coordinate information from global observation space by just using single vision equipment with one time static shooting.The technique of 3D reconstruction is one of the most important research goals of panoramic stereo spherical vision system.In this paper,we focus on discussing three modules of 3D reconstruction: calibration,epipolar rectification and stereo matching algorithm.The main works of this paper as follows:Panoramic Stereo Spherical Vision System was set up by fish-eye lenses,so in this paper we first discuss the mathematical model of fish-eye lenses.Based on it a geometric model and mathematical model of PSSVS is built,and then the parameters calibration method for PSSVS is realized.First,a general mathematical model which is applicable to the imaging law of fisheye lens is proposed and two distortion models are introduced: radial distortion model and tangential distortion model to get the distortion coordinates under the fisheye lens coordinate system.Then the distortion coordinates is projected on to the image coordinate system to obtain the complete mathematical model.Finally,the internal and external parameters of PSSVS are calculated by the mathematical model.Through the comparison of results,it shows that the proposed approach is rationality and validity.Due to the special construction of PSSVS,common epipolar rectification method is no longer suitable for stereo vision system.In this paper,through combining the spherical imaging model with epipolar constraint theory,the epipolar correction between the adjacent original fisheye images is achieved by using the system of internal and external parameters calibration results.First,according the projection formula,the image point is re-projected on to the sphere and then the sphere image is built.Then through rotation and translation,the ideal position of PSSVS is acquired.It makes the feature points and regions in same level and reduces the scope of matching to 1D.This method is suitable for fisheye lens imaging system whose optical is perpendicular.It reduces the complexity of the matching procedure and improves the matching accuracy to a large extent.In order to retain the intrinsic geometrical features of PSSVS,when realizing the stereo matching and 3D reconstruction,this paper proposed an algorithm based on maximally stable extremal regions(MSER)and affine scale invariant feature transform(ASIFT)without distortion correction first.The experiment results show that the algorithm can be applied in stereo matching of uncorrected fisheye image and the matching results meets the need of 3D reconstruction.