In this paper, we use a multilink-articulated wheeled in-pipe robot to develop an industrial solution for the visualized model of the small-size pipeline (less than 200mm) with unknown layout and diameter. We build visualized pipeline model based on small-size inertial measurement unit (IMU) and encoder instead of large-size optical device such as CCD cameras or radar which are expensive and complicated to operate easily. To improve the accuracy of the visualized pipeline model, a multi-sensor data fusion algorithm is developed and the error caused by the gravity factor has been eliminated by using the gradient descent algorithm. The proposed method is experimentally verified in U-Shaped pipeline.