The primary objective of this project is to mitigate the significant number of bicycle accidents that occur due to the cyclist’s inability to detect vehicles approaching from the rear. To address this, we have developed a system designed to heighten a cyclist’s awareness of potential road hazards. Drawing inspiration from the commonly used Mobileye collision avoidance system found in modern automobiles, our “Mobileye for Bicycles” provides the rider with crucial, real-time information about vehicles approaching from behind. The alerts generated by the system afford the cyclist additional time to react and implement measures to prevent an accident.
The system’s architecture is based on computer vision, integrating a YOLO model for object detection. This model is deployed on an OAK-D Lite camera, which handles on-device processing. A Raspberry Pi serves as the central control unit, running a sophisticated danger identification algorithm adapted from research by Mobileye. This controller also manages a dynamic database that logs vehicle detections, enabling the system to assess threats based on both historical and incoming data. Danger alerts are communicated to the cyclist through a wireless Bluetooth audio device. The entire system is designed to be portable and operate in a fully offline environment.
The results of this project confirm that it is possible to significantly improve a cyclist’s situational awareness through such a system. Nevertheless, the development process has identified numerous challenges that present opportunities for future research and system enhancement.