Using heuristic optimisation algorithms for camera calibration

Camera calibration can be defined as the determination of the geometric and optical properties of a camera and/or the 3D camera position and orientation with respect to a world coordinate system. Classical camera calibration techniques are based on mathematical models that utilize bundle blocks and employ the the bundle adjustment method. These models are linearized using the Taylor series expansion method to simplify complex equations. However, this linearization can lead to information loss and erroneous results. Additionally, classical methods tend to get stuck in local optima. On the other hand, heuristic optimization algorithms have the potential to reach global optima in numerical problems. These algorithms explore a broader search space without getting trapped in local optima, thereby offering several advantages over classical methods when used in camera calibration. In this thesis, the solutions of the well-known Tsai, Zhang models in the literature were compared with the Differential Search Algorithm (DSA), Artificial Bee Colony Algorithm (ABC), Particle Swarm Optimization Algorithm (PSO), and Multi-Population Differential Evolution Algorithm (MDE) methods. As a result of the research, it was seen that heuristic optimization algorithms are more successful than classical methods in camera calibration. These algorithms have provided more accurate and stable results without getting stuck in local solutions.