No description available.

This thesis describes a vision-based method for ROV¡¦s underwater pipeline recognition task. In this research, we tried to overcome the poor image quality of the underwater circumstance and the condition when seaweed is in the scene. Edge information and line feature of the pipeline are used in this method. Edge image is obtained after preprocessing to extract line feature. In this thesis we focused on the recognition of pipeline, trying to provide useful navigation information for further development of the ROV¡¦s control system.