Accident Prediction on E-Bikes Using Computer Vision

Micro Mobility refers to small, confined pace range, and lightweight modes of transportation, which include e-Bikes and e-Scooters. Micro Mobility riders are prone to harm and deadly injuries on city streets. Statistics show that on average 27 out of 82 (33%) humans, in the United States get injured every day in e-motorcycle injuries, which is rapidly increasing as compared to previous years. The current accident prediction strategies have been rather technically complicated to be incorporated onto micromobility vehicles which are lighter in weight and smaller in size. Current strategies bring key challenges to light. First, as micro-mobility vehicles are small and lightweight, the state-of-the-art accident prediction systems utilized in large vehicles such as cars cannot be directly applied. So extracting and providing insights has to be performed within real-time constraints and should be easy to install on micro-mobility vehicles. In addition, we are mostly restricted to the use of resource-constrained devices. The goal of this study is to develop fast and approximate accident prediction algorithms. This undertaking will take a look at approximate imaginative and prescient algorithms which includes optical flow with the drift, time to contact, and intensity estimation to extract actionable insights for a micro-mobility rider.