Comparison of radial distortion correction models for self photogrammetric camera calibration

In many fields such as computational vision, robotics and photogrammetry, the use of cameras is very important in order to execute different tasks. In or- der to achieve those tasks successfully, it is necessary to carry out a calibration of the cameras as an essential step before any of them. In photogrammetry, the reliability of camera calibration is essential to take decisive measurements. Camera lens distortions have a significant impact on image geometry and consequently, on camera calibration in general. The problem addressed in this thesis is the reliable radial distortion calibration of cameras, which is indispensable for being able to perform reliable measurements. This thesis compares the applications of different algebraic methods and their obtained models to the fully deterministic radial distortion correction only using 2 or- thophotographs of a personalized cube with a lattice pattern on it into. As remarkable results, it is experimentally proved that models obtained with the methods that minimizes the Orthogonal and Vertical distance always give best results for almost all the experiments. In addition , Wu’s method is the best interpolation method for a small feature point data set. Also, Cubic Spline has a little control over spikes which produces not very good results with bigger data sets for some experiments. Finally, clustering helps to reduce the feature points data set and also produces good radial distortion correction models.

• Автор: Molina Ron, María Fernanda (author)
• Формат: bachelorThesis
• Мова: eng
• Опубліковано: 2020
• Предмети: Photogrammetry; Camera Calibration; Radial lens distortion; Algebraic methods; Deterministic algorithms; Orthophotographs; Fotogrametría; Calibración de cámara; Distorsión radial del lente; Métodos algebraicos; Algoritmos deterministas; Ortofotos
• Онлайн доступ: http://repositorio.yachaytech.edu.ec/handle/123456789/218