Scholarlink

<p>The issue of solid wastes and contaminants in marine ecosystems has increasingly become a critical environmental challenge. In the previous study, our team focused on the impacts of floating plastic and how to capture them using innovative robotic solutions and categorize them into specific types. Building upon this foundation, we will broaden our scope to include a wider variety of underwater garbage. Driven by Artificial Intelligence (AI) and the Internet of Things (IoT), the primary goal of our robot Aqua Garbage Collector (AGC) is to monitor water pollution in real-time through camera modules by classifying and retrieving trash across various water depths. It will be programmed to send photos and coordinates of the detected trash to the central hub computer. Three machine learning models (YOLOv5, NanoDet, and RT-DETR) are studied to evaluate which model has the highest processing speed and accuracy. The results showed that YOLOv5 had the highest mean Average Precision (mAP), proving its real-time efficiency and accuracy. RT-DETR, while demonstrating good accuracy in complex cases, required longer training and showed lower mAP compared to YOLOv5. NanoDet, though highly efficient in computation, results in the lowest precision. The challenges of transmitting data underwater and over long distances are considered. Additionally, the model accuracy may be affected by the variability in water conditions, such as changes in turbidity and light penetration. This can result in a decrease in detection accuracy when operating in different underwater environments.</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

Aqua garbage collector: utilizing AI and IoT for efficient underwater garbage classification

  • 期刊论文

  • 发表日期：

    2025年6月16日

查看全文
  • Bonnie Yam,
  • Ruochen Zhao,
  • Maher Elshakankiri

摘要

The issue of solid wastes and contaminants in marine ecosystems has increasingly become a critical environmental challenge. In the previous study, our team focused on the impacts of floating plastic and how to capture them using innovative robotic solutions and categorize them into specific types. Building upon this foundation, we will broaden our scope to include a wider variety of underwater garbage. Driven by Artificial Intelligence (AI) and the Internet of Things (IoT), the primary goal of our robot Aqua Garbage Collector (AGC) is to monitor water pollution in real-time through camera modules by classifying and retrieving trash across various water depths. It will be programmed to send photos and coordinates of the detected trash to the central hub computer. Three machine learning models (YOLOv5, NanoDet, and RT-DETR) are studied to evaluate which model has the highest processing speed and accuracy. The results showed that YOLOv5 had the highest mean Average Precision (mAP), proving its real-time efficiency and accuracy. RT-DETR, while demonstrating good accuracy in complex cases, required longer training and showed lower mAP compared to YOLOv5. NanoDet, though highly efficient in computation, results in the lowest precision. The challenges of transmitting data underwater and over long distances are considered. Additionally, the model accuracy may be affected by the variability in water conditions, such as changes in turbidity and light penetration. This can result in a decrease in detection accuracy when operating in different underwater environments.