It is expected to apply AR (Augmented Reality) to support NPP (Nuclear Power Plant) maintenance for reducing human error and improving efficiency and safety. For the application of AR in plant maintenance work, real-time tracking technology with high accuracy is necessary. The objective of this study is to develop and evaluate a line feature-based tracking system that can be applied in NPPs for supporting fieldwork. In this study, the pose of a camera used as a sensor must be tracked in real-time. A PnL (Perspective n-Line) method is applied to realize the tracking. It calculates the position and orientation of the camera from the line correspondences between 3D-line landmarks and their projection on image. In order to make it possible to track in a capacious space, detected lines which are not registered as landmarks can be added as new landmarks if they are reliable line features. A performance experiment which evaluates the accuracy and running time of the system is being taken in lab environment. The result shows that the line feature-based method estimates the position and orientation of the camera with high accuracy in short distance tracking. Future experiment will be conducted in NPP to evaluation the feasibility of this system.