Abstract: With the increasing demand for digitalization and refinement in water conservancy engineering management, precise monitoring and management of reservoir dams are crucial. The traditional manual detection methods have many limitations, and the emergence of oblique photogrammetry technology has brought new opportunities for the 3D modeling of dams. This article takes Shengtian Reservoir in Bijie City, Guizhou Province as an example and uses oblique photogrammetry technology for 3D modeling. During the research process, the DJI "Yu" Mavic 2 Pro Professional Edition drone was selected to be equipped with a built-in high-resolution optical camera to collect data. The data was processed through techniques such as image distortion correction, feature extraction, regional network adjustment, and multi view image matching to construct a 3D real-scene model. The research results show that the constructed 3D real-world model has high geometric accuracy, with 97% of checkpoint distance errors less than 3 cm and an average distance error of 1.69 cm. It can fully restore the spatial geometric shape and surface texture details of the main structure and ancillary water conservancy facilities of the dam. Research has shown that the AFIST+SFM algorithm used in this article improves the reliability of the aerial triangulation point encryption method (referred to as "aerial triangulation encryption"), and the constructed model can provide intuitive visual data support for dam safety monitoring, operation and maintenance management, and water conservancy engineering planning. It also provides a reference technical path and engineering application paradigm for the digital modeling and structural safety assessment of similar reservoir dams.