Considering the issue of underwater pipeline recognition and tracking using autonomous underwater vehicle(AUV), the data flow is analyzed on the basis of simulation environment establishment, image processing, pipeline recognition and environment model. A complete underwater pipeline detecting and tracking system is proposed. In the process of pipeline recognition, to overcome the defect of line fitting using traditional Hough transform, an improved Hough transform is presented to dislodge the non-pipeline area in the view of feature clustering and to consider neighborhood of peak value process. Straight and curved pipelines are detected and tracked through the three-dimensional simulation environment for several times. Experiment results indicate that the presented system can effectively and autonomously detect and track pipelines with pipeline recognition and environment mapping.