Electronics, Journal Year: 2025, Volume and Issue: 14(10), P. 2065 - 2065

Published: May 20, 2025

To address the challenges of complex backgrounds, varying target scales, and dense targets in license plate detection within surveillance scenarios, we propose an enhanced method based on improved YOLOv8n network. This approach involves redesigning key components architecture, including C2f module, SPPF head. Additionally, optimize WIoU loss function, replacing original CIoU which leads to bounding box feature extraction regression accuracy. evaluate model’s robustness environments with lighting, angles, vehicle types, created a custom dataset. Experimental results show that model achieves notable increase accuracy, [email protected] rising from 90.9% baseline 94.4%, precision improving 90.2% 92.8%, recall increasing 82.9% 87.9%. parameters are reduced 3.1 M 2.1 M, significantly enhancing computational efficiency. Moreover, inference speed FPS 86, maintaining high meeting real-time requirements. demonstrates our provides efficient reliable solution for scenarios.

Language: Английский