Derailment Detection of Mining Shaft's Rail Vehicle Using Machine Vision on Edge Device

The railway system is pivotal in the mineral transportation industry. Due to rail track shape deformation or imbalance, the situation worsens. Stone, coal, and dust can also obstruct the mine rail transportation system. Moreover, derailment is very likely to happen, but conventionally, it is difficult to determine whether the vehicle is derailed. Scientists have developed different derailment detection techniques to overcome these difficulties, but most are implemented using sensors only. In challenging conditions, such as inside a mine shaft, sensor systems can fail for various reasons such as high dust levels, humidity, and fallen stones. These factors can easily destroy the sensors, especially if they are placed near to the ground surface. Machine vision, on the other hand, can detect objects from a greater distance than sensors. In our proposed system, we introduced a machine vision approach based on a Raspberry Pi, which combines YOLOv5n deep learning model and OpenCV to detect different types of derailment. Our system can be used as a rail track detection system by analyzing real-time video data.