The inspection management of special equipment is governed by national standards. The quality of special inspection reports directly impacts the authenticity, reliability, and national credibility of the inspection reports for special equipment. Due to the diverse types of special inspection equipment, complex inspection processes, and varying levels of inspection personnel, the standardization and reliability of inspection reports have not been well monitored. This paper develops an intelligent report sampling inspection system based on the Internet, deep learning intelligent core modules, front-end and back-end deployment, and MySQL database. The front-end is developed using VUE, and the back-end is deployed using SpringBoot. The design of the deep learning intelligent core modules includes a handwritten signature authenticity determination module based on improved GANomaly and a sampling result classification module based on HR-Net. Preliminary operational results of the system indicate that it can quickly obtain data from third-party systems, effectively determine signatures, and has a strong ability to distinguish text. The system significantly improves the inspection level of special equipment inspection reports.