Detection method for transparent window cleaning device, image processing approach.

Area of Science:

• Robotics and Automation
• Computer Vision
• Building Maintenance

Background:

• Increasing high-rise construction necessitates efficient exterior wall cleaning methods.
• Current exterior wall cleaning robots lack effective dust detection capabilities.
• Detecting dust on translucent glass, common in curtain walls, presents a significant challenge.

Purpose of the Study:

• To develop an image processing technique for detecting dust on translucent glass.
• To estimate the density of detected dust.
• To create a method suitable for outdoor conditions and adaptable to automated cleaning systems.

Main Methods:

• Utilized simple image processing techniques inspired by human visual perception.
• Employed a median filter for dust separation from the background.
• Applied mean shift analysis for dust density estimation.

Main Results:

• Successfully detected dust on translucent glass using image analysis.
• Estimated dust density effectively, even with blurry backgrounds.
• The method demonstrated robustness against variations in global brightness and background complexity.

Conclusions:

• The proposed image processing method offers a viable solution for detecting dust on building exteriors.
• This technique can be integrated into automated cleaning systems for high-rise buildings.
• The approach is practical for real-world outdoor applications due to its resilience to environmental factors.