3D Point Cloud Analysis for Enhanced Machine Learning-based Ultrasonic Inspection of Composite Materials

Description

An automated pipeline was presented for extracting and characterising the geometric features of a structure from high‐resolution 3D point‐cloud data, with the ultimate aim of discriminating real defects from false alarms in ultrasonic C-scan inspections. 3D point cloud-based geometrical feature extraction approaches were investigated using data measured by laser scanners, to enhance machine learning-based ultrasonic inspection of composite materials. Current ML-based ultrasonic C-scan images often lack the geometrical context required to differentiate between true defects and complex back wall features such as edges, ramps, or bends, limiting full automation in interpreting ultrasonic C-scan images.
This session introduces a technique that could fuse ultrasonic C-scan data with 3D geometrical context to improve reliability and NDT automation.

Period	31 Jul 2017
Event title	Lunch and Learn Programme at National Composite Centre (NCC): Lunch and Learn
Event type	Other
Location	Bristol, United KingdomShow on map
Degree of Recognition	Local