Abstract
Pipeline integrity and safety are critical for transporting water, oil and gas, but traditional inspection methods are resource-intensive and error-prone, delaying defect detection. The objective of this work is to create an autonomous robotic system capable of detecting, localizing and classifying problems inside a pipe using sophisticated imaging techniques and artificial intelligence algorithms. To achieve this goal, multimodal precision-enhancing sensors (high-resolution RGB cameras, ultrasonic and infrared sensors) with data processing methods such as the Canny edge detector and the DBSCAN clustering algorithm were used. The research approaches include the creation of a modular robotic platform for autonomous navigation, the creation of synthetic data for training deep neural networks, and experimental validation on pipelines of different materials and dimensions. The experimental results show that the system outperforms existing approaches, making it a valuable tool for predictive maintenance, regulatory compliance and improving pipeline safety.