Approach For Calibration And Orientation Of Digital Camera Based On Linear Features

To accurately reflect or make use of observation information, camera calibration, which is to correct distorted images obtained by camera and acquire interior orientation elements, has become an important component in measurement. And, camera calibration and orientation are prerequisite and basic problems for carrying on photogrammetry and computer vision, and lay a solid foundation for all aspects of application in the above fields.Due to the differences among the size of targets, measurement precision demands, and product types in computer vision and photogrammetry especially in close-range photogrammetry field, any calibration and orientation methods can not suit all applications. According to different applications and environments of images, it should apply to different methods. In addition, it is hard to propose a method to meet the demands of convenience, real-time and accuracy at the same time, especially in many cases of calibration camera parameters on the job are difficult to be recovered by geometrical relationships among sequence images. Consequently, excellent calibration and orientation algorithm is not only of great importance, but also a hotspot of study. In this paper, we firstly review the calibration method adopted in field. Because the images captured by photogrammetry and computer vision contains a mass of linear features, we have improved the calibration method from a single image, and proposed an approach for calibration of digital camera distortion based on non-control field. The exactness and feasibility in the theory and algorithm are proved by various experiments and analyses.Following aspects about these methods are introduced and focused on in this paper including:1. Approach for calibration of digital camera distortion based on non-control field.This paper proposes and demonstrates an approach based on the vanishing point theory for calibration of non-metric camera lens distortion. Images in this paper are photographed by Cannon-5d-35mm non-metric camera. The distortion correction by 3D control field,2D Control field and non-control field method respectively are studied experimentally. The results show that radial distortion plays an important role in optical distortion. The effects of radial distortion correction by the above mentioned methods are close to one another, and also prove the feasibility of calibration of non-metric camera distortion based on non-control field.2. Approach for camera calibration and orientation based on linear feature.The study on how to combine camera calibration with orientation to make both of them achieve better results is one of the emphases of this paper. In this paper, we acquire initial values of the camera calibration and orientation parameters using vanishing point theory and linear feature. Then, distortion parameters and a small amount control points are evaluated with self-calibration bundle adjustment. It will make camera calibration algorithm more stable and get higher accuracy.3. Make a summary of existing control fields.4. To compare diversified approaches for calibration of digital camera distortion and camera calibration and orientation. And analyze the influence degree of results on 3D measurement.Algorithms in this paper are especially suitable for 3D reconstruction of artificial structure. The architectures mostly contain the mutually perpendicular plane and linear features referred in the algorithms, so they are easy to operate and very flexible. The captured images can be used directly in calibration, orientation and 3D reconstruction. Therefore, they can economize on manpower, material and financial resources.