Camera Modelling and Calibration with Machine Vision Applications

Camera modelling and calibration are important parts of machine vision. They can be usedfor calculating geometric information from images. A camera model is a mathematical projectionbetween a 3D object space and a 2D image. The camera calibration is a mathematicalprocedure calculating parameters of the camera model, usually based on several images ofreference points. These fundamental parts of machine vision are improved in this thesis.One large part is the development of a generic camera model, GCM, that is accurate, computationallyefficient and can be used for both conventional, fisheye and even catadioptriccameras. Different models were used in the past for conventional and omnidirectional camerasand this is a well-known problem, the solution of which is described in this thesis. The accuracy of camera models is improved by introducing new ways of compensatingfor different distortions, such as radial distortion, varying entrance pupil point and decentringdistortion. Calibration is improved by introducing new means of calculating start estimates ofcamera parameters, from analysing shapes, sizes and positions of the reference points in theimages. These start estimates are needed in order to make the calibration converge. Methodsfor calculating better reference centre points than the centres of gravity are developed inorder to increase the accuracy further. Non-trivial null spaces that occur during calibrationare identified. Awareness of these improve the calibration. Calibrations with different camera models are implemented and tested for real camerasin order to compare their accuracy. Certain models are better for certain situations, but theoverall performance and properties are favourable for the GCM. A stereo vision weldingrobot system is developed, using the new model. It determines the geometry of a 3D weldjoint, so that a robot can follow it. The same system is implemented in a virtual environmentusing a simulation software. Such simulation is important since it makes it possible todevelop robot vision systems off-line