SegX-Net: A novel image segmentation approach for contrail detection using deep learning.

Area of Science:

• Environmental Science
• Computer Science
• Aerospace Engineering

Background:

• Aircraft contrails are significant contributors to global warming.
• Monitoring and mitigating contrail impact is crucial for climate change efforts.
• Current methods for contrail identification and monitoring face challenges.

Purpose of the Study:

• To propose and evaluate an advanced image segmentation technique for identifying and monitoring aircraft contrails.
• To introduce the SegX-Net architecture for efficient and accurate contrail segmentation.
• To provide a tool for the aviation industry to mitigate environmental impacts.

Main Methods:

• Development of the SegX-Net architecture, combining DeepLabV3+, upgraded, and ResNet-101.
• Evaluation of the model on a comprehensive dataset from Google research.
• Performance measurement using metrics like IoU, F1 score, Sensitivity, and Dice Coefficient.

Main Results:

• SegX-Net achieved superior segmentation accuracy.
• The model demonstrated an IoU score of 98.86% and an F1 score of 99.47%.
• Enhancements significantly improved the model's efficacy in contrail segmentation.

Conclusions:

• Image segmentation methods, like SegX-Net, show great potential for mitigating the climate impact of aircraft contrails.
• The SegX-Net architecture offers a viable solution for monitoring and managing contrail effects.
• This research contributes to the global fight against climate change through improved aviation environmental monitoring.