Scholarlink

This chapter introduces a scalable Smart Traffic Management System for real-time multi-camera surveillance in metropolitan environments. The system incorporates YOLOv11s for vehicle detection, OCR for license plate recognition, vehicle metadata extraction, and cross-camera reidentification with ResNet50-IBN-a with Milvus. Traffic infractions, such as red-light violations, wrong-way navigation, and accidents, are identified by rule-based spatial logic and orientation assessment. The system, implemented as FastAPI microservices, manages 50 camera streams at 15 FPS with a latency of 66 ms. Assessment indicates exceptional performance: 85.57% mAP for detection, 99.49% precision for license plate recognition, and 83.6% rank-1 accuracy for reidentification, underscoring significant real-world application.

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

A Scalable Smart Traffic Surveillance Framework Using Multi-camera Deep Vision

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Hoa Dam Nguyen Quynh,
  • Thong Vu Minh,
  • Anh Nguyen Hoang,
  • Long Nguyen Hai,
  • Nguyen Van Bay,
  • Duong Huu Thanh,
  • Nguyen Quoc Trung

摘要

This chapter introduces a scalable Smart Traffic Management System for real-time multi-camera surveillance in metropolitan environments. The system incorporates YOLOv11s for vehicle detection, OCR for license plate recognition, vehicle metadata extraction, and cross-camera reidentification with ResNet50-IBN-a with Milvus. Traffic infractions, such as red-light violations, wrong-way navigation, and accidents, are identified by rule-based spatial logic and orientation assessment. The system, implemented as FastAPI microservices, manages 50 camera streams at 15 FPS with a latency of 66 ms. Assessment indicates exceptional performance: 85.57% mAP for detection, 99.49% precision for license plate recognition, and 83.6% rank-1 accuracy for reidentification, underscoring significant real-world application.