A Safety Detection Method for Electric Bike Riding Incorporating License Plate Region Recognition

The widespread use of electric bikes has made rider safety a crucial issue in intelligent transportation research. To address the limitations of existing methods in terms of accuracy and robustness in electric bike license plate recognition, helmet detection, and rider detection tasks, this paper proposes a novel electric bike safety detection method, F-YOLOv8, which integrates license plate region recognition. Based on the YOLOv8 architecture, this method incorporates a feature pyramid network (FPN) to optimize multi-scale object detection capabilities, while introducing an attention mechanism to enhance the extraction of key features. The study constructs a high-quality dataset covering various riding scenarios and violations, including image data collected in Guangzhou and Foshan, Guangdong Province, annotated using LabelImg. The label types include Ebike (rider), plate (license plate), gzplace (license plate region), helmet (wearing a helmet), nohelmet (not wearing a helmet), and manned (carrying a passenger). Experimental results demonstrate significant improvements across all metrics: the precision on Ebike reached 96.1%, on Plate reached 92.9%, on Helmet reached 84.6%, and on Manned tasks reached 67.9%. This study provides an efficient and reliable detection method for enhancing electric bike riding safety management, laying a technical foundation for the development of intelligent transportation systems.