Vision System research for Autonomous Underwater Vehicle

A Vision System of an Autonomous Underwater Vehicle(AUVVS) is presented in this paper. The system combines computational vision and active vision. Image processing and object recognition, positioning, matching, these three aspects of research work were described in the paper. The first is underwater image processing, feature extraction and object recognition. Because of the property of underwater environment, several techniques were applied for image enhancement and segmentation. The second is concerned with matching. A matching algorithm based on building the map between image sequences is presented in the paper. The constraint of feature points connection is used to reduce search space. The third deals with recovering depth information and positioning the object. By the given lengths of edges, a new method for recovering 3D coordinate is implemented. Experimental results show the advantages of the system in robustness and real-time processing.