Learning-based computer vision approach for defect detection during robotic inspections

PI: Aparna Bharati

Co-PI(s): Mooi Choo Chuah

University: Lehigh University

Industry partner: FLX Solutions 

Every year, key infrastructure across the United States falls further into repair. There is a need to rapidly inspect, maintain, and upgrade these systems. FLX Solutions, Inc., a PA Keystone Innovation Zone, and Ben Franklin corporation, has developed a snake-like robot that enables workers to safely perform inspection and maintenance tasks in previously inaccessible areas using their 1-inch diameter FLX BOT robot. A key outcome of this PITA grant is the implementation of intelligent visual sensing algorithms that utilize machine learning-based algorithms for automatic anomaly and defect detection from videos recorded during infrastructure inspection. This is a critical need expressed by industry inspectors to increase the value of robotic inspections by accurate and efficient processing of inspection data. Methods for utilizing multiple cameras, like those found in every FLX BOT link, will also be explored allowing for improved visual understanding of the environment. This project is a collaboration between research, and industry through a partnership between the Computer Science and Engineering Department at Lehigh University and FLX Solutions, Inc. The results of the project will enable FLX Solutions to better address their customer needs through easy-to-implement detection algorithms allowing the FLX BOT to revolutionize the efficacy and efficiency of infrastructure inspection processes.