An Improved Generative Adversarial Network-Based and U-Shaped Transformer Method for Glass Curtain Crack Deblurring...

Area of Science:

• Computer Vision
• Artificial Intelligence
• Structural Health Monitoring

Background:

• Drones are vital for inspecting high-altitude glass curtain walls.
• Image quality degradation due to motion blur from vibrations hinders accurate crack detection.
• Existing deblurring methods may not be optimized for this specific application.

Purpose of the Study:

• To develop an advanced deblurring network tailored for drone-captured images of glass curtain cracks.
• To enhance the accuracy and reliability of automated crack detection systems.
• To introduce a specialized dataset for training and evaluating such systems.

Main Methods:

• A novel Generative Adversarial Network (GAN)-based and enhanced U-shaped Transformer network, named GlassCurtainCrackDeblurNet, was developed.
• A specialized dataset, GlassCurtainCrackDeblur Dataset, was meticulously created for this application.
• The proposed method was evaluated against established deblurring techniques.

Main Results:

• GlassCurtainCrackDeblurNet demonstrated superior qualitative and quantitative deblurring performance.
• The method effectively reduces motion blur in drone-captured images of glass curtain cracks.
• Improved image clarity directly benefits the accuracy of crack detection.

Conclusions:

• The proposed GlassCurtainCrackDeblurNet is highly effective for deblurring drone imagery of glass curtain cracks.
• This advancement significantly improves the potential for reliable, automated structural health monitoring of buildings.
• The specialized dataset facilitates further research and development in this domain.