Vision-Based Path Monitoring for Human Walkers - MSc Thesis Proposal by: Robert Odoh

Monday, August 11, 2025 - 13:00

The School of Computer Science is pleased to present…

 

Vision-Based Path Monitoring for Human Walkers

MSc Thesis Proposal by: Robert Odoh

 

Date: Monday, August 11th, 2025

Time: 1:00 pm

Location: Essex Hall, Room 122

 

Abstract:

This thesis introduces the Vision-Based Path Monitoring System (VB-PMS)—an assistive technology designed to enhance indoor navigation for visually impaired individuals using real-time computer vision. The system integrates YOLOv8, a state-of-the-art object detection model, with auditory feedback to detect obstacles and monitor veering from a straight walking path.

 

Addressing limitations in traditional mobility aids such as canes and guide dogs, the proposed method offers continuous environmental awareness, real-time obstacle alerts, and veer correction feedback using video analysis. The system captures input via a webcam or pre-recorded video, processes frames through a YOLOv8 pipeline to identify hazards, and translates visual detections into audio alerts using a lightweight text-to-speech engine.

 

Initial evaluations demonstrate smooth detection but reveal bottlenecks in inference speed and occasional false classifications (e.g., misidentifying suitcases). Ongoing enhancements include performance optimization, proximity-based bounding box filtering, and a user interface allowing toggling between video and live camera input. Ultimately, this research seeks to improve confidence, independence, and safety for users navigating complex, dynamic indoor environments.

 

Keywords: YOLOv8, Object Detection, Assistive Technology, Computer Vision, Audio Feedback, Indoor Navigation, Visual Impairment

 

Thesis Committee:

Internal Reader: Dr. Muhammad Asaduzzaman

External Reader:  Dr.  Severien Nkurunziza

Advisor:  Dr. Boubakeur Boufama