AI-enabled Interactive Threats Detection using a Multi-camera Stereo Vision System

Main Objective

Develop a vision-based inspection tool using stereo vision and Artificial Intelligence (AI) -enabled computer vision algorithms for pipeline threat detection and characterization. The proposed research consists of two major parts: hardware prototyping and software development. Hardware prototyping aims at design, calibration, and demonstration of the inspection device for pipeline in-line inspection. Software development focuses on developing deep learning-based algorithms for automated anomaly detection and physics-based models for evaluating possible interactions between detected threats.

Public Abstract

Pipeline infrastructure and safety are critical for sustaining a robust U.S. economy and for improving quality of life for the general public. Anomalies such as cracks and pits pose serious threats to the integrity of pipelines. As such, accurate and timely detection and characterization of anomalies is essential to support safe operation and integrity of U.S. pipeline systems. The proposed study aims to develop a novel multi-camera stereo vision system for in-line pipeline inspection; the design integrates hardware prototyping and software development for fast, accurate, automated pipeline anomaly detection. Physics-based models are developed to evaluate the effects of identified threats and their interactions on pipeline safety. The study also proposes a highly relevant educational component for training of next generation gas pipeline engineers through curriculum development and research experiences for undergraduate and graduate students. Potential technology transfer and commercialization of this system is planned through future collaborations and the establishment of an industrial advisory council.

Final Report

Final Report

693JK31950002CAAP_TEMPLATE_REPORT_01082023.pdf (6,229,718 bytes) [View] [Download/Save...]

Other Files

De-Brief Presentation

DeBrief_Presentation.pdf (7,931,028 bytes) [View] [Download/Save...]