Investigating the accuracy of crack detection using terrestrial laser scanner (TLS)

Abstract This study evaluated the accuracy of crack detection and width measurements under various scanning conditions. The data were collected through terrestrial laser scanner (TLS) scanning of the panel, which was tested with different scanning distances and incident angles, as well as cracks with known tilt angles on the panel. Mixed pixels are included in the point cloud import to eliminate the loss of valid crack information. A series of controlled experiments assessed how the inclusion or removal of mixed pixels affects the measurement accuracy. Gaussian fitting and intensity-based clustering were applied during data processing to enhance the precision of crack point extraction. The caliper-measured crack widths served as the ground truth, allowing the calculation of the absolute error percentage (AEP) for TLS-derived measurements. The results indicate that mixed pixel preservation in point cloud processing reduces AEP by 15%–30% compared to removal, while good scanning conditions −15°–30° incident angles, ⩽4 m distance, and ⩽30° tilt angles—optimize AEP to ⩽1.2% for 2 mm cracks and ⩽0.8% for cracks >2 mm. Larger angles (>60°) and greater distances (>6 m) increase the errors, particularly for cracks <4 mm. The proposed method is well-suited for laboratory testing, prefabrication quality control, and structural health monitoring tasks that require high-precision, non-contact crack measurement. It also provides practical guidance for the effective application of TLS in real-world engineering practices.