Thermal imaging for void detection and quantification in precast grouted structures using computer vision

Abstract

Grout penetration plays a pivotal role in ensuring the structural integrity and longevity of precast structural elements in construction industry. Insufficient grout penetration can lead to weak points in the structure, posing risks of leaks, collapses, and other forms of damage. This study analyzes the thermal images of grout-filled pipes to identify the areas of insufficient grout penetration using computer vision algorithms. A mathematical framework is developed to estimate unfilled grout volumes. The research compares the performance of pre-trained segmentation models and traditional image segmentation techniques, highlighting their effectiveness and limitations in detecting voids. Results demonstrate the superior accuracy of pre-trained models for anomaly detection, offering a promising approach for assessing the grout integrity in precast construction.

Keywords

• Thermal imaging
• Precast grout-filled pipes
• Computer vision
• Machine learning
• Image segmentation