Considering Spherical Refraction in Visual Ocean Gas Release Quantification

Abstract

Compared to traditional gas flow quantification methods, the stereo vision system has some advantages. However, underwater vision systems usually suffer from light refraction which can degrade the measurement accuracy from images. Cameras centered in spherical glass housings, dome ports, can theoretically avoid refraction, but misalignments in the dome create even more complex refraction effects than cameras behind flat glass windows. This paper introduces the spherical refraction model into a stereo vision gas flow quantification system. Also, this paper adds some contributions to an existing bubble quantification workflow for bubble size histogram and bubble volume estimation. First, the spherical glass dome port and the light propagation are modeled, and then the camera system is calibrated via underwater/in-air image pairs. Afterwards, the Epipolar Geometry Constraint is used to optimize the bubble matching. For volume estimation, an ellipsoid triangulation method is employed to improve ellipsoidal volume estimation. According to the calibration experiments and control experiments, the results show that the stereo vision gas flow quantification system can produce the volume of gas release accurately, which satisfies the requirements of long-term gas release monitoring in marine science.