Braille Block Detection via Multi-Objective Optimization from an Egocentric Viewpoint

Area of Science:

• Computer Vision
• Robotics
• Assistive Technology

Background:

• Visually impaired individuals rely on assistive devices for navigation.
• Accurate detection of tactile surfaces like Braille is crucial for these devices.
• Existing methods may not fully leverage diverse features for robust detection.

Purpose of the Study:

• To propose a novel method for detecting Braille blocks from an egocentric viewpoint.
• To enhance the functionality of walking support devices for the visually impaired.
• To address the challenge of robust Braille block detection in varied environments.

Main Methods:

• The task is framed as a multi-objective optimization problem.
• Both geometric and appearance features of Braille blocks are exploited.
• An evolutionary optimization framework is employed, modeling line pairs as solutions.
• Two objective functions are designed to clarify boundaries and estimate surface likelihood.

Main Results:

• The proposed method demonstrates effective detection of Braille blocks in real-world scenarios.
• Quantitative and qualitative results confirm the method's performance across various environments.
• A comprehensive comparison with different multi-objective optimization algorithms is provided.

Conclusions:

• The developed method offers a robust solution for Braille block detection.
• This advancement can significantly improve the capabilities of walking support devices.
• The multi-objective optimization approach provides a promising direction for future research in assistive technology.