Improve Nighttime Highway Vehicles and Pedestrian Detection Using Yolov8+CLAHE

The recognition of vehicles and pedestrians on roadways during nighttime is a significant difficulty within the domain of computer vision. Noise interference caused by inadequate illumination, shadows, and the glare of car headlights frequently impairs the effectiveness of object detectors. Aiming to boost the accuracy of nighttime vehicle and pedestrian identification, this work proposes combining the YOLOv8 algorithm with the Contrast Limited Adaptive Histogram Equalization (CLAHE) improvement approach. This work conducts a performance comparison of many YOLOv8 variations ($\mathbf{s}, \mathrm{n}, \mathrm{m}, \mathrm{l}, \mathrm{x}$) following the implementation of the CLAHE approach. The results indicate that the YOLOv8x variation exhibits a median Average Precision (mAP50) enhancement of $\mathbf{3. 9 \%}$ in comparison to the other.