With the widespread of the electric bikes in the urban traffic, it has become more and more important for the research on assessing the security and early warning system. This study aims to investigate safety assessment and early warning systems for electric bikes using head-mounted camera technology. By utilizing computer vision and deep learning techniques, the safety assessment and real-time warnings can be accurate and reliable. In this study, we design an environment perception algorithm based on head-mounted camera data. By analyzing the images from the rider's perspective, potential hazards such as pedestrians, vehicles, and other bikes can be detected and identified. In this process, deep learning algorithms are applied to train models that are able to understand and interpret the rider's field of view, accurately recognizing factors that may pose threats to electric bike safety. Finally, we integrate the safety assessment and behavior analysis results into a comprehensive early warning system. This system analyzes data captured by the head-mounted camera, continuously monitors the safety status of the rider, and issues alerts when potential dangers are detected, providing real-time safety information and guidance for electric bikes.