Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Effective Defect Detection Using Instance Segmentation for NDI

Ashiqur Rahman, Venkata Devesh Reddy Seethi, Austin Yunker, Zachary Kral, Rajkumar Kettimuthu, Hamed Alhoori

TL;DR

This work tackles defect detection in 3D ultrasonic scans of aerospace composite panels by casting the problem as instance segmentation on 2D projections, enabling direct localization of defects with minimal preprocessing. It evaluates two state-of-the-art frameworks, Mask-RCNN (Detectron2) and YOLOv11, on a COCO-formatted dataset derived from 3D ultrasound data, reporting competitive $mAP$ metrics at IOU thresholds $0.5$ and $0.75$ (e.g., $mAP^{50}=80.60\%$, $mAP^{75}=51.16\%$ for Detectron2; $mAP^{50}=77.04\%$, $mAP^{75}=48.41\%$ for YOLOv11). The results show that a lightweight, preprocessing-light pipeline can achieve effective defect localization with substantially reduced training times, particularly for YOLOv11, suggesting strong potential for real-time deployment in industrial NDI pipelines. The approach offers scalable adoption across varied scanning configurations, reducing inspector workload and enhancing safety by improving defect visibility and traceability.

Abstract

Ultrasonic testing is a common Non-Destructive Inspection (NDI) method used in aerospace manufacturing. However, the complexity and size of the ultrasonic scans make it challenging to identify defects through visual inspection or machine learning models. Using computer vision techniques to identify defects from ultrasonic scans is an evolving research area. In this study, we used instance segmentation to identify the presence of defects in the ultrasonic scan images of composite panels that are representative of real components manufactured in aerospace. We used two models based on Mask-RCNN (Detectron 2) and YOLO 11 respectively. Additionally, we implemented a simple statistical pre-processing technique that reduces the burden of requiring custom-tailored pre-processing techniques. Our study demonstrates the feasibility and effectiveness of using instance segmentation in the NDI pipeline by significantly reducing data pre-processing time, inspection time, and overall costs.

Effective Defect Detection Using Instance Segmentation for NDI

TL;DR

This work tackles defect detection in 3D ultrasonic scans of aerospace composite panels by casting the problem as instance segmentation on 2D projections, enabling direct localization of defects with minimal preprocessing. It evaluates two state-of-the-art frameworks, Mask-RCNN (Detectron2) and YOLOv11, on a COCO-formatted dataset derived from 3D ultrasound data, reporting competitive metrics at IOU thresholds and (e.g., , for Detectron2; , for YOLOv11). The results show that a lightweight, preprocessing-light pipeline can achieve effective defect localization with substantially reduced training times, particularly for YOLOv11, suggesting strong potential for real-time deployment in industrial NDI pipelines. The approach offers scalable adoption across varied scanning configurations, reducing inspector workload and enhancing safety by improving defect visibility and traceability.

Abstract

Ultrasonic testing is a common Non-Destructive Inspection (NDI) method used in aerospace manufacturing. However, the complexity and size of the ultrasonic scans make it challenging to identify defects through visual inspection or machine learning models. Using computer vision techniques to identify defects from ultrasonic scans is an evolving research area. In this study, we used instance segmentation to identify the presence of defects in the ultrasonic scan images of composite panels that are representative of real components manufactured in aerospace. We used two models based on Mask-RCNN (Detectron 2) and YOLO 11 respectively. Additionally, we implemented a simple statistical pre-processing technique that reduces the burden of requiring custom-tailored pre-processing techniques. Our study demonstrates the feasibility and effectiveness of using instance segmentation in the NDI pipeline by significantly reducing data pre-processing time, inspection time, and overall costs.
Paper Structure (17 sections, 2 figures, 2 tables)

This paper contains 17 sections, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. NDI.
  4. NDI Techniques.
  5. ML in NDI.
  6. Image Segmentation.
  7. Data
  8. Method
  9. Mask R-CNN (Detectron $2$)
  10. Training Configuration.
  11. Results.
  12. YOLO v11
  13. Training Configuration.
  14. Results.
  15. Discussion
  16. ...and 2 more sections

Figures (2)

  • Figure 1: Sample of signals showing the difference between defect and non-defect areas
  • Figure 2: Exported images of the ultrasonic scans. A) The exported image used for the training. B) Orange labels showing the defect location on the image. C) Defects identified by the Detectron $2$ model displayed with green labels. D) Defects identified by the YOLO $11$ model displayed with blue labels.