Improved detection of subsurface defects through active thermography and ensembling techniques

Abstract

Quality control and defect detection are major challenges in industrial environments. The localization of these defects is of crucial importance, as they can affect the performance of manufactured products and even the safety of people. Defect detection methods based on visual sensors and image processing are nowadays the most common approaches. To detect subsurface defects it is common to combine active thermography and deep learning. In this paper, PCT (Principal Components Thermography) is employed to enhance the SNR (signal-to-noise ratio), leading to a significant improvement in the results. Recent developments in this field apply deep learning models for semantic segmentation and object detection. However, the quality of the predictions is not enough to ensure that all quality controls are met. In this paper, a combination of semantic segmentation and object detection is proposed to increase the reliability of predictions. To carry out this combination, a working methodology and two new ensembling strategies are proposed. After comparing the results of the proposed combination with the results of using only semantic segmentation methods in a real industrial scenario, where carbon fiber sheets are used, it is found that the proposal improves the segmentation metrics by a 5% to a 24%. Thus the reliability of the predictions is improved.

Keywords:

• model ensembling
• object detection
• quality control
• semantic segmentation
• subsurface defect detection

Additional information

Funding

This research was funded by the project PID2021-124383OB-I00 of the Spanish National Plan for Research, Development and Innovation.

Notes on contributors

Darío G. Lema

Darío G. Lema received the M.S. degree in computer science from University of Oviedo, Spain, in 2021. He is currently a Ph.D. Candidate at the University of Oviedo. In recent years, he has been working on several projects related to computer vision and industrial systems. His research is based on image processing, thermography and real time systems.

Oscar D. Pedrayes

Oscar D. Pedrayes graduated from the M.S degree in Computer Science at the University of Oviedo, Spain, in 2021. He is currently pursuing his PhD in Computer Science at the University of Oviedo doing research in the field of computer vision, in particular, in convolutional neural networks (CNNs) for semantic segmentation. Some of his research areas include thermography, real-time systems, cloud systems, defect classification, crop classification and satellite image classification, among others.

Rubén Usamentiaga

Rubén Usamentiaga is a Full Professor in the Department of Computer Science and Engineering at the University of Oviedo. He received his M.S. and Ph.D. degrees in Computer Science from the University of Oviedo in 1999 and 2005 with the Extraordinary Doctorate Award. His research interests include machine vision, real-time imaging systems and infrared thermography, where he has published more than eighty papers in JCR journals with five Prize Paper Awards. He has been a visiting professor at the Universities of Laval and Pennsylvania. He is also a senior member of the IEEE and the IAS society.

Daniel F. García

Daniel F. García is a full professor in the Department of Computer Science and Engineering at the University of Oviedo, Spain, where he leads the area of computer engineering. His current research interest is in the area of self-adaptive computer systems, including cloud computing, IoT systems and services, and computer vision systems. During the last 25 years, he has lead many research projects and coauthored 102 papers in journals and more than 150 communications in conferences and workshops. He is a member of the IEEE Computer Society.