Adaptive calibration of an underwater robot vision system based on hemispherical optics

In this paper the issue of on-line adaptation of the robot vision system to variations in the refractive index of the ambient fluid is being considered. This is achieved by developing an analytical model of the light propagation through the AUV water-dome-air interface before entering the camera optics. The model developed is used for simulation of the AUV's camera dome. The effect of variations in the refractive index of the ambient fluid is investigated through experimental and simulation results. Based on this model, an appropriately designed stationary target is mounted on the AUV in the camera's FOV enabling adaptive estimation of the fluid refractive index. Experiments were carried out in and out of the water, thus changing the refractive index of the ambient fluid. Relations of the system's sensitivity to the camera resolution and the target geometry are investigated. The refractive index measurements obtained through the developed on-line adaptive estimator during the experiment, closely match the ones predicted by the model through the simulations. The experiments were performed using a commercially available ROV.