Improve the safety of e-scooters for young commuters based on sensors and object detection

Abstract: The increasing popularity of electric scooters in urban areas has coincided with a significant rise in traffic accidents involving these vehicles. Although deep learning-based object detection has been effectively applied to improve automobile safety, its application for enhancing electric scooter safety remains underexplored. This study presents an advanced obstacle-detection system specifically designed for electric scooters. The system integrates an RGB camera, depth camera, inertial measurement unit (IMU), and a Jetson Nano board to enable efficient real-time processing. The deep learning-based object detection model identifies common road hazards such as tree branches, manhole covers, and potholes. By incorporating depth data, the system accurately assesses obstacle proximity and issues timely alerts to riders, facilitating proactive hazard avoidance. The IMU monitors ground vibrations, providing essential ground truth data that enhances the reliability of obstacle detection. This comprehensive approach, which combines computer vision, edge computing, and data fusion, ensures real-time hazard awareness and improves e-scooter safety. The system’s ability to detect and communicate potential road infrastructure hazards enhances risk mitigation, contributing to a safer urban mobility experience. The integration of object detection, depth analysis, and IMU-based vibration monitoring underscores the system’s effectiveness, establishing it as a robust tool for real-time safety improvements tailored to electric scooter users.