Scholarlink

With rising concerns about global warming, the focus is shifting toward renewable energy, particularly solar power. Calculating the area of solar panels is crucial for determining energy output. This task is efficiently accomplished through satellite image segmentation. Earlier methods relied on complex neural networks such as Convolutional Neural Networks and Fully Convolutional Networks, which improved object detection but struggled with scalability and retaining details. This research leveraged efficient and cost-effective methods for measuring solar panel areas using architectures such as Feature Pyramid Network, DeepLabV3, Pyramid Scene Parsing Network, UNet, and UNet++. Various encoders are used to assess segmentation accuracy against ground truth through metrics like Intersection over Union, Precision, and Dice Similarity Coefficient. This study has shown that the UNet architecture with the EfficientNetB7 encoder achieves the highest accuracy, with an IoU of 0.90 and a Dice Coefficient of 0.94. This combination demonstrates exceptional precision and reliability in segmenting solar panel areas, affirming its effectiveness in the analysis.

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

Optimized Segmentation for Accurate Rooftop Solar Power Estimation Using Satellite Images

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Bandam Deekshitha,
  • Paduru Akshaya Reddy,
  • Krushi Sirimalle,
  • T. Satyanarayana Murthy,
  • K. Gangadhara Rao,
  • P. Ramesh Babu

摘要

With rising concerns about global warming, the focus is shifting toward renewable energy, particularly solar power. Calculating the area of solar panels is crucial for determining energy output. This task is efficiently accomplished through satellite image segmentation. Earlier methods relied on complex neural networks such as Convolutional Neural Networks and Fully Convolutional Networks, which improved object detection but struggled with scalability and retaining details. This research leveraged efficient and cost-effective methods for measuring solar panel areas using architectures such as Feature Pyramid Network, DeepLabV3, Pyramid Scene Parsing Network, UNet, and UNet++. Various encoders are used to assess segmentation accuracy against ground truth through metrics like Intersection over Union, Precision, and Dice Similarity Coefficient. This study has shown that the UNet architecture with the EfficientNetB7 encoder achieves the highest accuracy, with an IoU of 0.90 and a Dice Coefficient of 0.94. This combination demonstrates exceptional precision and reliability in segmenting solar panel areas, affirming its effectiveness in the analysis.