Extraction and Calculation of Roadway Area from Satellite Images Using Improved Deep Learning Model and...

Area of Science:

• Remote Sensing
• Urban Planning
• Geospatial Analysis

Background:

• Roadway area calculation is a novel challenge in remote sensing and urban planning.
• Accurate road extraction from satellite imagery is crucial for urban development and management.

Purpose of the Study:

• To propose an effective method for calculating roadway area from satellite images.
• To enhance the accuracy of roadway extraction using deep learning architectures.

Main Methods:

• A two-step approach: roadway extraction followed by area calculation.
• Utilizing novel U-net++ and ResNeXt convolutional neural network architectures.
• Implementing an efficient post-processing technique to improve overlap with ground truth data.

Main Results:

• The proposed method achieves superior performance in road extraction compared to existing U-net family models.
• The approach demonstrates high accuracy in calculating roadway area using pixel resolution from satellite images.
• Evaluation on the Massachusetts dataset confirms the effectiveness of the novel architecture and post-processing.

Conclusions:

• The developed method offers a significant advancement in automated roadway area calculation.
• This technique provides a robust solution for urban planning and remote sensing applications.
• The combination of U-net++ and ResNeXt architectures presents a state-of-the-art solution for road extraction.