To solve the problems of low efficiency, strong subjectivity, and prone to missed inspection in quality and safety hazards inspection during the construction of the Yangtze River protection water conservancy projects, the specific scenarios of various quality and safety hazards were identified by analyzing the requirements of multi-scenario quality and safety inspection of the projects. The YOLOv5 algorithm was used for image enhancement and optimization, and an intelligent recognition algorithm framework was established. Through on-site photography, web crawling technology, and internal data resources of the project department, thousands of high-quality photos were collected and organized to obtain a comprehensive dataset. On this basis, by integrating the regional inspection functions, the quality and safety inspection system can accurately monitor only the designated work areas. This effectively avoids false inspection and improves inspection efficiency and accuracy.