Design of a Computer Vision-based Third-party Risk Alert System for Oil & Gas Pipeline Corridors

This paper presents the development of a platform designed to detect and manage third-party risks in pipeline corridors. The project involved the collection of aerial images of pipeline areas, featuring various vehicles that pose potential risks. Using this dataset, we trained YOLOv9 models to identify these vehicles. Our evaluation identified the YOLOv9-C model as the most effective, with a precision of 86.6%, a recall of 74.5%, and a mean Average Precision of 83.1%. The chosen model was then incorporated into a custom-built platform designed to manage detections and alerts. This platform features components for visualization, logging, tracking, and notification dispatch. It demonstrates how integrating Artificial Intelligence into asset monitoring operations can greatly improve the efficiency and accuracy of risk detection and management.