Design of state inspection system for small targets on railway wagons based on video detection

With the development of new technologies such as artificial intelligence and big data, the railroad freight industry has developed to a new level, and the safety condition of wagons, as the carrier of railroad transportation, has been more widely concerned. Limited by the data quality and algorithm effectiveness, the automatic state detection has poor monitoring accuracy of small targets. In this paper, a state inspection system for small targets on railway wagons was developed. The fusion SPD-Conv layer was used to improve the yolov8 model to realize the rope loosening detection. A context-aware network module was employed to replace the fusion layer that comes with yolov8x to realize the open wagon hitch abnormality detection. The mean average precision of rope belt loosening detection reaches 0.873, and the precision for detecting wagon hitch abnormalities reaches 0.948, proving the system's effectiveness in practice.