An Intelligent Method for Train Accident Prevention and Detection Using ResNet and Long Short-Term Memory Network

At unattended railroad crossings, numerous technologies have been developed to monitor trains, identify objects, and automatically open and close gates. In order to prevent accidents at the crossing, the currently planned systems may effectively close and open the gates upon the arrival and departure of trains. However, there are frequently incidents at railroad crossings caused by the unexpected arrival of trains. Indulging in headphone use is a leading cause of accidents on railroad tracks. The literature review revealed that no viable approach for preventing such mishaps had been developed. The purpose of this proposed system is to provide a comprehensive analysis of the current systems dedicated to the prevention of accidents in real time. Preprocessing, feature selection, and model performance evaluation are all used in the suggested method. The DWT method is used for preprocessing in the suggested method. PCA and LDA are used for feature selection. The suggested method employs ResNet LS TM to assess the efficiency of the model. Two conventional approaches, LSTM and GRU, are contrasted with the proposed methodology. The proposed method outperforms the other two.