INTERNATIONAL JOURNAL OF ADVANCE RESEARCH
IN MULTIDISCIPLINARY

Manholes, essential components of urban utility infrastructure, provide access to underground systems like sewers, electrical conduits, and storm drains. However, deteriorated, open, or missing manhole covers pose significant hazards to pedestrians, cyclists, and vehicles, often leading to severe accidents. Traditional inspection methods rely on manual observation, which is labour-intensive, error-prone, and inefficient, particularly in large urban areas. Moreover, the increasing frequency of manhole-related incidents highlights the urgent need for an automated and reliable solution to ensure public safety and efficient maintenance of urban infrastructure. This project addresses these challenges by proposing an advanced deep learning based automated inspection system. The system utilizes Convolutional Neural Networks (CNN) for image classification and You Only Look Once version 8 (YOLOv8) for accurate detection and localization. It is trained on a diverse dataset to classify manhole covers into distinct categories, including 'Closed,' 'Open,' 'Broken,' 'Overflow,' and 'No Manhole.' The integration of UAV images and CCTV footage ensures comprehensive monitoring, even in hard-to-access areas or dynamic environments. By overcoming issues like variable image quality and complex backgrounds, this solution offers precise and timely identification of hazardous conditions. The implementation of this system presents a transformative approach to urban safety and maintenance. By automating the inspection process, it reduces reliance on manual labour, minimizes errors, and ensures timely intervention to address potential risks. This project not only enhances public safety but also optimizes resource allocation for infrastructure maintenance, offering a scalable and efficient solution to modern urban challenges.