Real-Time Accident Detection and Reporting System Using Edge Computing and Convolutional Neural Networks with Integration of GPS for Public Safety

Accidents on roads are a significant cause of injuries and fatalities worldwide. The majority of the vehicles on the roads are not adequately equipped with sensors to detect accidents. Detecting accidents in such vehicles is a challenge. This paper proposes a system for the timely detection of accidents using Convolutional Neural Networks (CNN) from the vehicles passing by to provide immediate medical attention in the time of need to reduce the impact of the accident. This study discussed both deep learning and machine learning models to detect accidents. The proposed system includes cameras and Global Positioning System (GPS) sensor interfaced with edge devices like Nvidia Jetson Nano to enable edge computing and is mounted on vehicles. The proposed system features automatic notification about the accident to the emergency services and first responders, along with the precise location of the accident. This paper also proposes a web page to validate and identify false predictions by the model and update the dataset to improve the system’s efficiency in future versions. The proposed system’s performance is evaluated on a publicly available dataset. Machine Learning like Support Vector Machine (SVM), K-Nearest Neighbour (KNN), Random Forest, and Deep Learning models including CNN, ResNet50, and InceptionV3 are employed for accident prediction. A parametric analysis of accuracy, precision, and recall is performed to effectively classify the best techniques. The CNN model proves to be effective and is used as the model to interface with the Jetson nano board.