Camera Calibration Based on Self-Adaptive Cuckoo Search Algorithm

Abstract

Camera calibration is one of the most important processes in computer vision. To obtain more robust results of optimization calibration, an improved self-adaptive cuckoo search algorithm for camera calibration is proposed in this paper. The step length formula is set adaptively so that the objective function can avoid local minima and miss the optimum solution. What's more, the improved cuckoo search algorithms used in non-linear camera calibration do not need initial estimation values. So the proposed method can solve the problem that traditional optimization algorithms are sensitive to initial value. Furthermore, the self-adaptive cuckoo search algorithm combined with the process of camera calibration is used to optimize the camera's intrinsic parameters and the coefficient of radial distortion. Finally, the average re-projection error is analyzed, and the mean absolute error and the standard deviation are also calculated on the cases of different noise level. The experimental evaluation demonstrates that the proposed method is considerably more accurate and robust.