Accurate Underwater Camera Calibration Method by Non-Parametric Model for Distortions Correction

Light is refracted when traveling through water, glass and air, and the underwater images are distorted. In this case, it is not feasible to use the traditional pinhole imaging model to characterize underwater imaging directly. An accurate calibration method for underwater three-dimensional (3D) reconstruction is proposed based on an optical mixed distortion model and the bundle adjustment algorithm. Firstly, the nonlinear imaging model of underwater optical system is constructed by analyzing the imaging distortion errors caused by underwater optical refraction. Then, an improved self-calibration algorithm with two-step bundle adjustment is developed to calculate the internal and external parameters of underwater imaging system, fully considering the errors caused by refraction and distortion. Finally, a hybrid-optics experimental system for 3D underwater reconstruction is developed to verify the performance of the method. Experimental results demonstrate that the standard deviation of the four-camera system is 0.222mm with field view of 1.2m×0.8m×0.5m, and the quality of the reconstructed point clouds is accurate enough for many underwater 3D imaging fields, so the proposed method is feasible for underwater three-dimensional reconstruction.