Scholarlink

To ensure the structural integrity of submerged marine concrete structures, early crack detection and localization are crucial. Traditional methods like ultrasonic pulse velocity (UPV), dye penetration, and diver checks often face challenges such as poor visibility, high costs, and safety risks. This paper presents an advanced Remotely Operated Vehicle (ROV) system for efficient crack detection. The ROV uses a Raspberry Pi 5 and ESP32 microcontroller for processing and control. A 5MP camera captures real-time video processed with the YOLOv11 model for crack detection, enhanced by 20W COB lights for visibility. Directional control is managed remotely via a Firebase API. The system includes a floatation device housing a GPS module and Wi-Fi router for location tracking and communication. Real-time video feeds are sent to a GPU equipped ground station, where cracks are localized accurately. Incremental learning improves detection by saving detected crack images and retraining the model to adapt to varying underwater conditions thereby solving the issue of inadequate dataset of underwater crack images. This system offers a reliable and safer solution for underwater inspections.

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

Intelligent Underwater ROV for Real-Time Crack Detection in Marine Structures

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • P. B. Nithin,
  • Deepa Elizabeth George,
  • Roshni Polly,
  • Vaishnavi Balasubramanian,
  • Aiswarya James

摘要

To ensure the structural integrity of submerged marine concrete structures, early crack detection and localization are crucial. Traditional methods like ultrasonic pulse velocity (UPV), dye penetration, and diver checks often face challenges such as poor visibility, high costs, and safety risks. This paper presents an advanced Remotely Operated Vehicle (ROV) system for efficient crack detection. The ROV uses a Raspberry Pi 5 and ESP32 microcontroller for processing and control. A 5MP camera captures real-time video processed with the YOLOv11 model for crack detection, enhanced by 20W COB lights for visibility. Directional control is managed remotely via a Firebase API. The system includes a floatation device housing a GPS module and Wi-Fi router for location tracking and communication. Real-time video feeds are sent to a GPU equipped ground station, where cracks are localized accurately. Incremental learning improves detection by saving detected crack images and retraining the model to adapt to varying underwater conditions thereby solving the issue of inadequate dataset of underwater crack images. This system offers a reliable and safer solution for underwater inspections.