A Method Of 3D Reconstruction Based On PAL Cameras

3D reconstruction based on images is aimed at obtaining 3D model of the object from a sequence of images. It is one of the means to solve the 3D modeling problem in the field of computer vision, and it has great practical significance for real time 3D modeling of large-scale and complex scenes. The research results currently have been applied in the field of medical imaging, autonomous navigation and digital entertainment etc. However, there are still many problems in the process of 3D reconstruction, such as complex multiple cameras calibration, difficult 3D modeling etc. To solve the problems above, in this paper, the method of 3D reconstruction based on panoramic cameras is proposed, including deploying and correcting the panoramic image, matching the image features and calculating 3D coordinates of points. The main contents of this paper are as follows:Firstly, this paper uses the panoramic camera to capture images. The image captured by the panoramic camera is an annular image. The position relationship in the annular image is analyzed to calculate the center, the inside diameter and the outside diameter of the annular image. The optical properties of the PAL camera are determined by the least square method through shooting the wall calibrated with equal distance. The annular image is transformed into the rectangular image, and the distortion in the transformed image is corrected by the optical properties of the PAL camera.Secondly, 3D reconstruction method proposed in this paper just requires simply camera calibration, which greatly simplifies the reconstruction procedure. According to the characteristics of the panoramic image, the position lines in the panoramic image are matched as the features. Two panoramic cameras vertically placed take two panoramic images. Dynamic programming algorithm is taken to match the two images. The stretching correction and the adjacent distance correction are added to improve the matching effect.Finally, according to the triangulation principle, the horizontal distance from the camera to the object is obtained. The geometric relationship between the corresponding point in space and the corresponding points in two images is deduced by the position relationship in the cylindrical coordinate. Then 3D coordinates of the points in the image can been calculated.Experimental results show that: the method proposed in this paper is simple, effective, and the error between the reconstruction model and the actual model is less than 10 cm.