Research On Technologies Of Digital Close-Rang Industrial Photogrammetry Based On Trifocal Tensor

Digital close-range industrial photogrammetry is kind of measurement method that makes use of more photos taken from different direction and the conversion between the various coordinate systems and the calculation of the corresponding. It includes a number of key steps, such as the design the encored targets were designed and decoded, camera calibration and the three dimensional reconstruction. The main content in this paper is the analysis and research of two parts: camera calibration of digital close-range photogrammetry system and three-dimensional reconstruction of the test target point. In digital close-range industrial photogrammetry, the common calibration modes have the problem of the large limitations of calibration plate and manually match. In order to solve these restrictions, a new camera calibration method based on orientation circle was proposed, and the automatic detection and matching of target points was achieved and the high-precision camera parameters were obtained and optimized by the Gauss-Newton. Because of low precision in three-dimensional reconstruction applying two images, the research in this paper got rid of the shackles of traditional photogrammetric methods, combined with trifocal geometric principles in computer vision and used trifocal tensor to complete the stereo matching of the images and to obtain the three-dimensional coordinate values of the test points. The main contents and achievements are as follows:Firstly, on the basis of the analysis of the calibration principles and restriction in traditional calibration plate camera calibration mode, a new high-precision digital camera calibration plate and its model based on circle target points was designed. In the new calibration plate, according to the geometric and the positional information of the five orientation circles on the calibration plate, the linear relationship among the respective directions based on cost function model was built and the automatic detection and matching of all target points was achieved. Minimize nonlinear equations was built, combined with the Gauss-Newton to solve the optimal values of camera’s inside and outside parameters. The experiments in Matlab2010 programing environment were conducted and the results indicated the correctness and applicability of the proposed theory in this paper.Secondly, coding target points were used as the matching primitive and trifocal tensor was used as epipolar line constraint, the image stereo matching was achieved to solve the problem of low precision in correct three-dimensional reconstruction. Under the condition that the camera motion and scene structure were not constrained, accurate three-dimensional reconstruction was achieved applying three images shot by aforementioned calibration camera.Finally, According to the geometric properties and algebraic properties of trifocal tensor, the constraints of the cameras inside and outside parameters and the coupling relationship of known structure information in the scene, the three-dimensional coordinates of the target point was finally solved out and the three-dimensional reconstruction of the scene was achieved. In order to reduce the computational error and noise impact, the average distance of all space points and the algebraic relations of scale factors were studied and the scale factors was solved out and the dimension was unified. Through the experiments, three-dimensional reconstruction based on trifocal tensor proposed in this paper had higher accuracy and was more suitable for the application in practice.