Accident Detection and Reporting System for Powered Two-Wheelers Using Internet of Things and Edge AI augmented with MCUSUM features

According to data provided by the Ministry of Road Transport and Highways (MORTH), Government of India, about 1.68 lakh people have lost their lives in accidents and nearly 4.4 lakh people have suffered injuries [1]. The number of accidents has increased by approximately 12% year on year. The data also show that people affected by accidents are likely to be in the age group of 15 to 49 years. This age group is virtually the pillar that supports the nation's development. Therefore, road fatalities can indirectly influence the nations development. In this work, we have implemented a system aimed towards powered two-wheelers (PTWs) that detects accidents and sends an SOS signal to a preconfigured contact. The system, upon detecting an accident/critical event, can attempt to convey the SOS signal in multiple modalities for redundancy. The SOS signal includes the last known GNSS coordinates of the vehicle to be able to pinpoint the location of the incident. The system also has a manual SOS trigger option that can be used as a panic button. Detection is performed by a Convolutional Neural Network (CNN) based classifier running on the system using Edge AI capability that uses sensory output from an inertial measurement unit (IMU) and its derivative data for input features. In this work, we have experimented with the possibility of improving the classification performance of the classifier by augmenting the input feature set with Multivariate Cumulative Summation (MCUSUM) [2] control chart statistic of the IMU data. The results show that this technique can give marked improvements in the overall accuracy of the classification system.