YOLOv10-Driven Enhanced Vehicle Detection in Low-Light On-Board Environments
N.Jathin Reddy, department of Computer Science and Engineering, GNITC, 22-5L1, 22wj1a05l1@gniindia.org
L.R.P.L.Poojitha, department of Computer Science and Engineering, GNITC, 23-518,23wj5a0518@gniindia.org
M.Shashikanth, department of Computer Science and Engineering, GNITC, 21-5J6,21wj1a05j6@gniindia.org
Mohd.Irfan, Assistant Professor, department of Computer Science and Engineering, GNITC
Abstract - Accurate vehicle detection in low-light environments remains a significant challenge in intelligent transportation systems and autonomous driving applications. Vision-based on-board detection systems often experience degraded performance under poor illumination conditions due to factors such as motion blur, headlight glare, and increased visual noise. To overcome these challenges, this research proposes a YOLOv10-driven enhanced vehicle detection framework designed for real-time deployment in low-light on-board environments. The proposed approach utilizes the advanced feature extraction capabilities and optimized architecture of YOLOv10 to improve detection accuracy, computational efficiency, and robustness compared with earlier object detection models. In the proposed framework, several image pre-processing techniques—including adaptive histogram equalization, noise reduction, and contrast enhancement—are applied to improve the visibility and quality of input images before they are processed by the detection model. The YOLOv10 model is further fine-tuned using a diverse dataset consisting of nighttime and low-illumination driving scenarios, enabling the system to generalize effectively under challenging environmental conditions. The lightweight yet powerful design of YOLOv10 supports real-time inference on embedded and edge devices, making it highly suitable for in-vehicle applications where processing speed and resource efficiency are essential. Experimental results demonstrate that the proposed YOLOv10-based detection model achieves superior performance in terms of mean Average Precision (mAP), detection speed, and reduction of false positives when compared with baseline models such as YOLOv8 and Faster R-CNN. Furthermore, the system shows strong resilience to common low-light challenges including shadow occlusion, glare from artificial lighting, and environmental noise, thereby ensuring reliable vehicle detection across a wide range of nighttime driving conditions.
Key Words: YOLOv10, Vehicle Detection, Low-Light Image Enhancement,Computer Vision,Intelligent Transportation Systems, Night-Time Vehicle Detection
