Pillay, NSingh, NSivate, Themba Mthembane2026-06-222026-06-222024https://hdl.handle.net/10321/6409Submitted in fulfilment of the requirements for the degree Master in Engineering: Computer Engineering, Durban University of Technology, Durban, South Africa, 2024.A primary challenge faced by individuals with visual impairments is the difficulty or inability to perform object identification. While conventional aids such as magnifying spectacles may assist with near-field vision, individuals with reduced or absent sight often rely extensively on tactile exploration for object recognition. This reliance presents significant navigational challenges, particularly in dynamic environments such as roadways, where the increased risk of accidents is a major concern. To address this, this research proposes the development of a portable machine vision system designed to provide real-time auditory feedback regarding detected proximal objects, thereby assisting individuals with visual impairments in navigation. The design of the system prioritizes portability, reliability, modularity, and unobtrusiveness during typical operation. The hardware implementation of the proposed system consists of three key elements: a Single Board Computer (SBC), a wireless camera, and a Bluetooth-enabled earpiece. The experimental results demonstrate that the proposed system is capable of delivering real- time audio feedback of detected objects to visually impaired individuals. The system was evaluated in a real-life environment in both acceptable and poor lighting conditions. The efficacy of the proposed system in well-lit environments resulted in an average detection rate of 87.65%. However, in low-light scenes an average detection rate of 51% was observed because of the low image resolution. The minimum observed delay was 4.9 seconds, while the maximum was 10.2 seconds. This latency encompasses the duration required for image capture, processing, and audio translation. The latency can be mitigated by incorporating an integrated circuit with a dedicated Graphical Processing Unit (GPU), which is more proficient in handling machine learning and video processing tasks.122 penComputer visionObject detectionAssistive technologyVisual impairmentBlind navigationMachine learningEmbedded systemsReal-time processingAudio feedbackWearable computingArtificial intelligenceAccessibility technologyComputer visionMachine learningArtificial intelligenceComputers and people with disabilitiesSelf-help devices for people with disabilitiesThesishttps://hdl.handle.net/10321/6409