Wagon and Container Codes Detection and Recognition based on YOLOv8

Railway systems have become an essential and growing element for freight transport sector due to the low levels of pollution it offers compared to other alternatives. However, they have not seen a corresponding advancement in technology, as many of the inspection procedures are still based on outdated technologies or even performed manually. This work proposes a method for automating the reading of UIC (wagons) and BIC/ILV (containers) codes based on YOLOv8, both for detection and recognition processes. The challenge is that, unlike in the case of car license plates, codes on freight trains can be located in different parts of the train or even appear fragmented on different parts of the container, present irregular lighting conditions, or have different formats due to the lack of strict standardization. The proposed method has been compared using real images obtained from a railway operator, achieving significantly better results compared to previous works.