Building surface defect detection is a key step in performing building restoration, building stability and safety testing. However, since building facades are characterized by complex environments (light, shading, etc.), single-image detection techniques may suffer from inaccurate detection results, despite being fast. Moreover, the results of defect detection can provide data and technical support for the construction of the subsequent building operation and maintenance platform.
This study presents an analysis framework that combines image and point cloud data to improve the accuracy and efficiency of building surface defect detection. On one hand, the region of interest (ROI) is extracted from the image, and a fusion of the local binary pattern (LBP) and Otsu's method is employed to segment the target from the background. On the other hand, after preprocessing the point cloud data corresponding to the ROI, the alpha shapes algorithm is applied to detect building cracks, while random sample consensus (RANSAC) plane fitting is utilized to identify concave-convex deformations on the building surface.
The results show that the image can detect the architectural surface defects, but the results are too much to distinguish the type of surface defects; the laser point cloud can detect the extraction of cracks and concave-convex, and the results are in line with the actual situation, and the combination of both can improve the accuracy of the detection results while ensuring the detection efficiency.
More data, especially point cloud data, needs to be collected for training to get more accurate results. Spalling and bulging on a building’s external wall may result from the construction process, and further distinction between architectural decoration and structural defects is needed.
The study paves the way for developing an integrated operations and maintenance (O&M) management platform by combining defect detection with BIM technology. Leveraging image and point cloud data of buildings or components, the platform generates comprehensive defect records, including defect location, type, size, severity, and BIM ID. This enables the creation of a holistic O&M management system that seamlessly integrates safety, cost, and maintenance considerations.
Reduce misjudgment by distinguishing cracks and spalls with image texture (LBP + OTSU) and point cloud data (alpha shapes + RANSAC) combined with the physical rules of defects (deviation and area). And it can achieve millimeter-level defect detection. Rapidly obtain and identify the current condition of building facades to provide a basis for building damage assessment.
