Extraction and Analysis of Blue Steel Roofs Information Based on CNN Using Gaofen-2 Imageries

Area of Science:

• Remote Sensing
• Urban Planning
• Environmental Science

Background:

• Blue steel roofs are cost-effective and durable, commonly found in industrial areas.
• Accurate mapping of blue steel roofs is crucial for assessing industrial inefficiency and urban heat islands.
• Industrial areas are vital for manufacturing, necessitating efficient spatial analysis.

Purpose of the Study:

• To accurately map and analyze the quantity and spatial distribution of blue steel roofs in Foshan's Nanhai district.
• To evaluate the performance of the DeeplabV3+ semantic segmentation model for blue steel roof detection.
• To investigate the correlation between blue steel roof area and industrial gross output.

Main Methods:

• Utilized GaoFen-2 satellite imagery for analysis.
• Applied the DeeplabV3+ semantic segmentation neural network.
• Compared DeeplabV3+ performance against maximum likelihood classification.

Main Results:

• DeeplabV3+ achieved an overall accuracy of 92%, outperforming maximum likelihood classification.
• Blue steel roof distribution varied across the study area but was uniform at the town scale.
• A significant positive correlation was found between the area of blue steel roofs and industrial gross output.

Conclusions:

• The DeeplabV3+ model is effective for mapping blue steel roofs, providing high accuracy.
• Findings support the identification of inefficient industrial areas for regional planning.
• Data generated is fundamental for urban environmental studies, including heat island research.