研究人员开发了一种新方法，以改进实时车牌检测和识别 (LPDR) 系统。该方法通过引入跨空间混合注意力 (CSHA) 和类别平衡合成增强 (CBSA) 来解决训练集中的空间字符不匹配和数据不平衡问题。这项技术将较少见车牌的识别率从 78.2% 显著提高到 91.5%，同时保持了 152 FPS 的高处理速度。 AI
影响 通过提高车牌识别系统的准确性和效率，增强智慧城市中的实时人工智能应用。
排序理由 该集群包含一篇详细介绍车牌识别新方法的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2606.05785v1 Announce Type: cross Abstract: Real-Time License Plate Detection and Recognition (LPDR) forms the backbone of modern smart cities. Although the YOLOV5-PDLPR model substantially improved system efficiency through a parallel decoder approach, its performance is s…
Real-Time License Plate Detection and Recognition (LPDR) forms the backbone of modern smart cities. Although the YOLOV5-PDLPR model substantially improved system efficiency through a parallel decoder approach, its performance is still affected by spatial character mismatches and …