Identifying the Defects Presented on the Exterior Layers of a Structure by Employing 3D Point Clouds and Thermography

Abstract

Severe weathers and climate changes do have profound impacts on the exterior layer of structures. Especially for those defects presented on the exterior layers of buildings, those defects can make the exterior layers of buildings worse, and, sometimes, the life cycles of buildings are decreased. How to identify those defects shown on the exterior layers of structures is an important issue for the sustainable structure management. Non-destructive testing (NDT) methods have been widely applied to detect those defects presented on the A terrestrial layers. However, NDT methods cannot provide efficient and reliable information for identifying the defects because of the huge examination areas. Thermography is a product of a thermal infrared camera. Thermography is used to record the surface temperature information, and the differences of the recorded surface temperatures are small such that it was difficult to analyze the collected thermographs. The defects presented on the exterior layer of buildings are usually located at the places whose surface temperatures are higher than their corresponding neighbors. This paper proposed to employ image segmentation to cluster those pixels with similar surface temperatures such that the thermography can be composed of the limited groups. For each segmented group, the surface temperature distribution in each segmented group is homogenous. In doing so, the regional boundaries of the segmented regions can be identified and extracted. A terrestrial laser scanner (TLS) is widely used to collect the point clouds of a structure, and those point clouds can be modeled as a 3D structure. This study established a mapping model such that the segmented thermography can be projected on the 3D structure. In doing so, the 3D structure provides the defects and their spatial locations. In this paper, the structure (a wind turbine) located at Taichung County Gaomei Wetlands is used as an example.