Adebanjo, HannahAdeliyi, Timothy T.Mthimkhulu, Zinhle2026-06-262026-06-262026-03-16https://hdl.handle.net/10321/6419A dissertation submitted in fulfilment of the requirement for the Master’s in Information and Communications Technology degree, Durban University of Technology, Durban, South Africa, 2026.The oil and gas (O&G) sector plays a critical role in sustaining global energy demands, yet pipeline deterioration, particularly corrosion, remains a persistent threat to operational safety and service continuity. While several robotic inspection systems have emerged over the years, most remain prohibitively expensive and inaccessible, leading many South African oil and gas (O&G) companies to rely on slow, traditional inspection methods. This study responds to this gap by designing and developing a portable, and cost-effective in-pipe inspection robot that integrates image acquisition, sensor technologies, and both classical and deep-learning-based corrosion-detection techniques. Guided by a comprehensive review of existing inspection technologies, the research identified key limitations relating to cost, accessibility, and operational constraints, which informed the design requirements of the proposed robot. Lowcost components were systematically selected and integrated into a lightweight platform capable of navigating pipelines of varying diameters using a four-wheel configuration supported by two omnidirectional wheels. A high-resolution camera and sensor suite were incorporated to enable reliable image capture, environmental awareness, and obstacle avoidance. A combination of image processing algorithms, Canny Edge Detection, Sobel Operator, K-Means Clustering, and a Convolutional Neural Network (CNN) was implemented to detect corrosion with improved accuracy. Functional testing demonstrated stable mobility, effective sensor communication, and extended battery performance, validating the robot’s operational feasibility despite initial challenges related to low-light image quality, which were mitigated by upgrading to a higher-resolution camera. The findings confirm that the proposed in-pipe inspection robot can reliably navigate pipelines, acquire and process images, and detect corrosion, offering a practical and economically viable alternative for pipeline inspection in the O&G sector. Overall, the study established a foundation for accessible, scalable, and technology-driven corrosion-monitoring solutions. The research contributes to the field of pipeline inspection by presenting a validated, cost-effective in-pipe inspection robot that improves access to automated corrosion detection and supports safer and more efficient infrastructure maintenance in the O&G sector. The findings show that reliable automated pipeline corrosion detection can be achieved using an affordable robotic approach, thereby enhancing the practical feasibility of deploying smart inspection technologies in resource-constrained operational environments.139 penIn-pipe inspection robotFrugal roboticsPipeline inspectionPipeline corrosionCorrosion monitoringImage processingCanny edge detectionSobel operatorK-means clusteringConvolutional neural network (CNN)Embedded systemsLow-cost roboticsSmart inspectionOil and gas pipelinesSouth AfricaPipelines--CorrosionPipelinesRobots--Design and constructionDeep learning (Machine learning)Computer visionNondestructive testingThesishttps://doi.org/10.51415/10321/6419