为实现高质量、高效率的隧道巡检，基于AR技术设计隧道人工巡检辅助系统。通过AR终端使得真实场景叠加虚拟信息，对巡检作业进行分析、指导和决策，有效辅助隧道人工巡检工作。应用SURF的尺度不变和高效的识别性能对机电设备进行识别匹配，并采用Lowe’s算法保证匹配的最佳效果；通过隧道5G室内高精定位、物体的POI信息、AR智能终端中的方向角度传感器等数据信息，建立基于目标物体唯一属性的识别匹配；最后，选择PSO信息优化算法对SURF算法识别匹配的结果和唯一属性识别匹配的结果进行优化关联融合，极大地提升识别匹配的准确性。从匹配性能、AR渲染性能并结合实际运行效果对AR隧道虚拟巡检系统和传统人工巡检效果做出对比评估，验证了AR巡检辅助系统相对于人工巡检有明显优势。系统所展示的设备实时数据及辅助决策功能使巡检平均耗时缩减为传统人工巡检的54%，按AR辅助巡检系统规划的巡检路线可有效避免漏检的情况，且巡检总体准确率接近94%，有效提升了巡检质量与效率。

 A tunnel manual inspection auxiliary system is designed based on augmented reality(AR) technology to achieve highquality and highefficiency tunnel inspection. The virtual data are applied to the real scenario using the AR terminal to analyze, guide, and make decisions on tunnel inspection, thus effectively assisting manual tunnel inspection. The scale invariance and efficient recognition performance of speededup robust features(SURF) are used to identify and match electromechanical equipment. Lowe′s algorithm is used to ensure the best matching effect. Recognition and matching based on the unique properties of the target object are established using data from tunnel 5G indoor highprecision positioning, object′s point of interface information, AR intelligent terminal direction, and angle sensor. Finally, the particle swarm optimization information and algorithm are used to improve the correlation and fusion of the SURF algorithm and unique attribute recognition and matching results, which considerably improve the recognition accuracy and matching. The actual operation effect is compared with the traditional manual inspection effect from the matching performance and AR rendering performance perspectives, confirming that the AR inspection auxiliary system has certain advantages over manual inspection. Furthermore, the realtime equipment data and decisionmaking functions displayed in the system reduce the average inspection time by 46% compared to the traditional manual inspection. Thus, the inspection route planned by AR auxiliary inspection system can effectively avoid missing inspections. The overall inspection accuracy is approximately 94%, which effectively improves the quality and efficiency of the inspection.