基于数字孪生技术，深入探讨建筑信息化模型（building information modeling，BIM）与智能图像识别技术在智能巡检领域的集成挑战及其对泵站设备异常等异常事件快速响应能力的提升方法。将BIM轻量化后与网络地理信息系统（Web geographic information system，WebGIS）相结合，为三维智能巡检搭建数字孪生场景。在此基础上，集成视频融合技术与智能识别算法，构建BIM技术耦合图像识别技术的三维智能巡检系统，实现BIM与视频监控虚实交互，快速直观地展示告警信息，辅助值班人员快速响应设备异常等事件。结果表明：BIM技术耦合图像识别技术在数字孪生泵站中具有巨大的潜力，可以有效提高巡检质量，为泵站智慧化管理作依托。

The application of digital twin technology in the South-to-North Water Transfers Project is explored, with a specific emphasis on a pioneering digital twin project at a pumping station, aiming to promote the digitization trend in water resources engineering. Focused on the construction goal of the digital twin pumping station, which aims to achieve "unmanned and minimally manned" operations, the research highlights the critical transition from sensory analysis to intelligent analysis for the realization of key intelligent capabilities at the pumping station. While advanced methods such as unmanned aerial vehicle intelligent surveillance, smart image recognition technology, and intelligent inspection robots have been applied in the field of intelligent substations, limitations exist in the visualization of monitoring results.To fully consider the multifactorial impacts on the operational safety and reliability of the pumping station, the need for the intelligent inspection of the digital twin pumping station is emphasized, combining Building Information Modeling for precise modeling, Web Geographic Information System technology, and monitoring videos to provide more accurate visual feedback. The research adopts the integration of Building Information Modeling and Web Geographic Information System technologies to construct a three-dimensional scene. The three-dimensional video integration technology is used to ensures the accurate overlay of monitoring videos onto the geographical coordinates of the three-dimensional model. Through the Cesium engine, the coordinates of monitoring cameras are standardized to World Geodetic System 1984, achieving precise integration of virtual scenes with real-time monitoring. Simultaneously, with the linkage between Building Information Modeling models and monitoring, event listeners are defined to dynamically call monitoring platform interfaces, detecting changes in camera perspectives and ensuring synchronization with the scene. In terms of intelligent image recognition, real-time recognition of numerical values on equipment dials and indicator light statuses is achieved, combined with monitoring device data to judge equipment abnormalities.Significant practical results have been achieved through the research. The system is equipped with high-resolution monitoring devices, enabling automatic triggering of alarms upon detecting anomalies through real-time data collection and image processing. Furthermore, the digital twin pumping station's three-dimensional intelligent inspection system efficiently generates detailed inspection reports, encompassing inspection results, monitoring images, sensor data, and equipment status information. In application scenarios, the system automatically focuses on problematic areas, highlighting abnormal equipment and achieving digitalized intelligent inspections, thereby enhancing equipment and environmental safety and reducing potential risks.In summary, a comprehensive technical solution for the three-dimensional intelligent inspection system of the digital twin pumping station is provided. Despite facing some challenges, such as improving three-dimensional video integration effects and image recognition results, efficient operations for the digital twin pumping station have been achieved through the introduction of strategies like progressive Building Information Modeling model lightweighting. Future directions include further advancing three-dimensional video integration and intelligent image recognition technologies to enhance visualization and inspection quality, promoting collaborative operations among components, and realizing unmanned operations for the pumping station's intelligent management on a more robust foundation.