Efficient Measurement of Structural Defect Depth Using Parallel Laser Line-Camera System

Abstract

The precise depth measurement of common structural defects, such as bulging, delamination, and spalling, is paramount in building condition assessment. This paper presents an efficient and portable parallel laser line-camera system designed for accurately reconstructing defect depth profiles from projected laser stripes. The system features a telescopic design to enhance the measurement range and operational flexibility. Central to its efficacy is a machine learning–aided image processing algorithm that facilitates both robust and highly accurate depth measurements. Specifically, advanced deep learning techniques are applied to detect and segment laser stripes from background interference. A novel hypothesis optimization (HO) algorithm, grounded in a three-layer backpropagation (BP) neural network, is proposed to reduce errors in laser baseline recovery caused by image distortion further. Comprehensive laboratory and field experiments validate the measurement accuracy and superior noise suppression capabilities of the system. Additionally, the paper studies potential errors that could emerge during field operations, thereby confirming the practical utility of the device. The proposed system quickly generates surface profiles in a single shot, making it a valuable tool for monitoring uneven objects.