Aircraft PFD information recognition system based on YOLOv5

The Primary Flight Display (PFD) plays a pivotal role in providing pilots with real-time access to critical flight parameters and is essential for flight training documentation and operational status monitoring. To address the limitations of traditional video recording methods—particularly their inefficiency in storage utilization and information extraction—this study proposes a PFD information recognition system based on the YOLOv5 object detection framework. The system begins by constructing a high-quality PFD image dataset and employs a YOLOv5 model to automatically detect key labels such as airspeed, altitude, and heading. A spatial logic inference mechanism is then introduced to intelligently interpret flight states such as climb, descent, and stall. The system's front end is developed using the PyQt5 framework, supporting the import of images and videos, real-time recognition, status prompts, and structured data export. Experimental results demonstrate that the system achieves high recognition accuracy and real-time performance across various flight conditions. Moreover, by converting raw video data into lightweight, structured spreadsheet files, the system maintains information integrity while significantly improving data usability and storage efficiency. The results indicate strong engineering applicability and significant potential for broader adoption.