Abstract: The paper aims to empower indoor mobile terminals' fast and accurate navigation based on computer vision. Based on the DeepTag system used in indoor scenarios, the usability of different fiducial markers is evaluated. The paper studies the positioning principle of DeepTag system and tests the recognition accuracy of traditional fiducial markers in indoor practical application scenarios. According to the experimental results, it is concluded that ArUco markers is the most suitable for the DeepTag system and best fiducial marker type posted on walls in indoor scenes, which can provide a reference for the augmented technology of AR navigation in indoor environment based on computer vision in the future.