Keypoint-based ROI detection for vehicular optical camera communication

Optical camera communication (OCC) for vehicle-to-everything (V2X) scenarios has emerged as a critical technical pathway in intelligent transportation systems (ITS). However, the coupled interference between background noise and communication signals in complex outdoor environments severely compromises the OCC system’s robustness, necessitating precise region-of-interest (ROI) detection algorithms for light source localization to ensure reliable information transmission. To solve this problem, this study proposes a keypoint-based ROI (KP-ROI) detection model and establishes a systematic evaluation framework for V2X communication performance. Integrated with OCC system architecture, this framework defines the detection rate (${\alpha }_d$) and precision rate (${\alpha }_p$) to analyze ROI’s impact mechanisms on communication quality while employing Undersampled Differential Phase Shift On-Off Keying (UDPSOOK) modulation for experimental validation. The experimental results show that in controlled static scenarios, the proposed algorithm can stably transmit vehicle communication data at a communication distance of 15 m with an average bit error rate (BER) of 0.5%, and supports an angular deflection of ±30°. In dynamic scenarios, during straight-line constant-speed driving, the BER can remain stable below 5%. This study verifies the strong correlation between ROI detection robustness and communication efficiency, while demonstrating the advantages of the keypoint algorithm in light source localization tasks.