This paper is the result of geological survey engineering.

Objective Manual geological hazard inspection suffers from low operational efficiency, numerous constraints and high operational risks. Therefore, it is of great significance to study how to realize efficient and accurate identification of geological hazards in large areas.

Methods This study focuses on the geologically hazard-prone areas of Xi'an as the research site. It develops a methodology for interpreting geological hazards in large areas using real-scene 3D models. By employing aerial photogrammetry, a high-precision real-scene 3D model covering 6,004.2 km² of the study area was obtained. Based on an analysis of the developmental characteristics of geological hazards in the area, interpretation markers were established for various types of hazards, including collapses, landslides, and debris flows. Through a phased, iterative human-computer interaction process involving "sample area interpretation followed by on-site verification," the approach circumvents the problems of low interpretation accuracy and substantial ineffective on-site verification efforts associated with one-time comprehensive interpretation of large areas. This method effectively enhances interpretation accuracy and facilitates the efficient and precise identification of geological hazards across thousands of square kilometers.

Results Ultimately, 546 previously undetected geological hazard sites were identified, with an overall interpretation accuracy rate of 88.35%. This confirms the feasibility of the interpretation methodology and the effectiveness of the interpretation markers. The efficiency of geological hazard inspection significantly surpasses that of traditional manual survey methods.

Conclusions Practical applications have demonstrated that the research findings can substantially improve the accuracy and efficiency of geological hazard identification, offering a novel technical approach and valuable reference for the efficient inspection of large-scale geological hazards across extensive regions.