Simple vision tracking of pipelines for an autonomous underwater vehicle

A 2D vision based pipeline tracking system is described for an autonomous underwater vehicle (AUV). A model-based recognition approach is used. From a 2D camera image rectangles are extracted. The real-world size and direction of each rectangle are estimated with the aid of input from a sonar and a compass. The rectangle that is consistent with the diameter and local direction of the pipeline is used to guide the AUV along the pipeline.<<ETX>>