Yolov10-Driven Enhanced Vehicle Detection In Low-Light On-Board Environments

Abstract

Accurate detection of vehicles under low-light conditions remains a significant challenge in intelligent transportation systems and autonomous driving applications. Vision-based onboard systems frequently experience performance degradation due to poor illumination, motion-induced blur, headlight glare, and increased image noise. This paper proposes a real-time vehicle detection framework based on YOLOv10, specifically designed for low-light onboard environments.The proposed approach incorporates image enhancement techniques, including adaptive histogram equalization, denoising, and contrast optimization, to improve visual quality prior to detection. The enhanced images are then processed using a fine-tuned YOLOv10 model trained on a diverse dataset containing nighttime and low-visibility driving scenarios. This enables improved generalization and robustness under challenging lighting conditions.YOLOv10’s optimized architecture provides a balance between accuracy and computational efficiency, making it suitable for deployment on resource-constrained embedded systems. Experimental results demonstrate that the proposed framework outperforms baseline models such as YOLOv8 and Faster R-CNN in terms of mean Average Precision (mAP), inference speed, and reduction of false detections.

Furthermore, the system shows strong resilience against common low-light challenges, including shadow interference, artificial light glare, and environmental noise. These results indicate that the proposed method is a reliable and efficient solution for real-time vehicle detection in nighttime and low-illumination driving scenarios.