Scholarlink

<p>This work describes an end-to-end License Plate Recognition (LPR) framework that can handle challenging weather conditions such as haze, rain, mist, and insufficient light. The pipeline combines contemporary image preprocessing methods such as Contrast-Limited Adaptive Histogram Equalization (CLAHE) and morphological filtering, a UNet-based dehazing module based on residual learning expansion, and the latest state-of-the-art deep learning architecture for detection and recognition. YOLOv8 is used for license plate localization, while PaddleOCR is employed for high-precision character recognition. The dehazing module significantly improves image quality by eliminating atmospheric noise, resulting in improved detection and recognition performance. Experiments on synthetic weather-degraded datasets show the robust performance of the system, with recognition accuracies of 95% under clear weather and up to 90% under fog and mist. Comparative and ablation experiments also confirm the effectiveness of each pipeline component. This paper demonstrates a highly robust and modular approach to License Plate Recognition (LPR) suitable for real-world traffic enforcement and monitoring applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

U-Net- and YOLOv8-based pipeline for license plate recognition in adverse weather

  • 期刊论文

  • 发表日期：

    2025年11月10日

查看全文
  • Pranav Padmakumar,
  • Edwin Lisho,
  • Esther Moriah A,
  • Johanan Joysingh S

摘要

This work describes an end-to-end License Plate Recognition (LPR) framework that can handle challenging weather conditions such as haze, rain, mist, and insufficient light. The pipeline combines contemporary image preprocessing methods such as Contrast-Limited Adaptive Histogram Equalization (CLAHE) and morphological filtering, a UNet-based dehazing module based on residual learning expansion, and the latest state-of-the-art deep learning architecture for detection and recognition. YOLOv8 is used for license plate localization, while PaddleOCR is employed for high-precision character recognition. The dehazing module significantly improves image quality by eliminating atmospheric noise, resulting in improved detection and recognition performance. Experiments on synthetic weather-degraded datasets show the robust performance of the system, with recognition accuracies of 95% under clear weather and up to 90% under fog and mist. Comparative and ablation experiments also confirm the effectiveness of each pipeline component. This paper demonstrates a highly robust and modular approach to License Plate Recognition (LPR) suitable for real-world traffic enforcement and monitoring applications.