Scholarlink

To develop an automated system for detecting tailings ponds, a significant environmental concern due to their toxic contents. Current methods, including manual inspections and satellite imagery, are time-consuming and lack real-time data. The proposed system leverages deep convolutional neural network algorithms and remote sensing imagery to improve detection accuracy and efficiency. The project’s ultimate goal is to enhance environmental monitoring capabilities and enable early identification of potential risks. The training accuracy of pretrained convolutional neural networks (CNN) models InceptionV3, visual geometry group (VGG16), and ResnetRS152 is 99.7%, 97.57%, and 95.54%, and the validation accuracy is 95.28%, 95.3%, and 78.3% for classification, respectively. The mean average precision mAP (50) of the object detection model you look only once version 5 (YOLOv5) is 71.6%, and YOLOv8 is 67.3%. It represents a significant step forward in the field of ecological monitoring. By integrating advanced technologies and developing an automated system, it aims to address the issues of tailings pond detection and monitoring effectively.

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

Advancing Environmental Monitoring: Deep Learning Techniques for Tailings Pond Detection and Categorization

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Zin Din Ziden,
  • Pulak Islam,
  • Zeshan Ahmed,
  • Ishtiaq Hoque Farabi,
  • Syeda Aynul Karim,
  • K. M. Safin Kamal,
  • Ahmed Wasif Reza

摘要

To develop an automated system for detecting tailings ponds, a significant environmental concern due to their toxic contents. Current methods, including manual inspections and satellite imagery, are time-consuming and lack real-time data. The proposed system leverages deep convolutional neural network algorithms and remote sensing imagery to improve detection accuracy and efficiency. The project’s ultimate goal is to enhance environmental monitoring capabilities and enable early identification of potential risks. The training accuracy of pretrained convolutional neural networks (CNN) models InceptionV3, visual geometry group (VGG16), and ResnetRS152 is 99.7%, 97.57%, and 95.54%, and the validation accuracy is 95.28%, 95.3%, and 78.3% for classification, respectively. The mean average precision mAP (50) of the object detection model you look only once version 5 (YOLOv5) is 71.6%, and YOLOv8 is 67.3%. It represents a significant step forward in the field of ecological monitoring. By integrating advanced technologies and developing an automated system, it aims to address the issues of tailings pond detection and monitoring effectively.