An Indoor Obstacle Detection System Using Depth Information and Region Growth

Area of Science:

• Computer Vision
• Robotics
• Assistive Technology

Background:

• Visually impaired individuals face challenges navigating unfamiliar environments.
• Existing obstacle detection systems may struggle with ground plane segmentation and over-segmentation issues.

Purpose of the Study:

• To develop an effective obstacle detection system using depth information for the visually impaired.
• To address and overcome the over-segmentation problem in ground plane removal.
• To enable safer navigation for visually impaired individuals in new surroundings.

Main Methods:

• A three-part system: scene detection, obstacle detection, and vocal announcement.
• A novel ground plane removal technique using the Connected Component Method (CCM).
• CCM addresses edge and initial seed position problems in region growth methods.

Main Results:

• The system successfully detects both static and dynamic obstacles.
• The proposed ground plane removal method overcomes over-segmentation.
• Experimental results demonstrate the system's robustness and convenience.

Conclusions:

• The developed system provides a simple, robust, and efficient solution for obstacle detection.
• This technology can significantly improve the independence and safety of visually impaired individuals.
• The method's effectiveness in handling ground plane segmentation is a key advancement.