<![CDATA[Failure to identify traffic signs triggers road accidents, demanding autonomous vehicle navigation systems robust against lighting dynamics. Addressing the lack of objective measurements in previous studies, this research aims to design and evaluate a YOLOv11-based detection system on a miniature autonomous car prototype. The methodology includes hardware design (ESP32, webcam, actuators) and model training using 12,455 augmented images. Performance was evaluated quantitatively via real-time testing under controlled lighting (1–300 Lux) using a lux meter, then analyzed with third-order polynomial regression to prove non-linear correlations. Computationally, the model achieved an mAP@0.5 of 99.43%. Field testing demonstrated high robustness at 1 Lux confidence score (0.77), peaking at 50 Lux (0.87), but dropping drastically at 300 Lux (0.03) due to lens glare. Conclusively, YOLOv11 is highly reliable in normal lighting, but its implementation requires Lateral Overflow Integration Capacitor (LOFIC) optical sensors to overcome extreme overexposure.]]>
Copyrights © 2026
