Identification of Transparent and Specular Reflective Glass Planes in TLS Data with Intensity Values

Abstract

When setting up a terrestrial laser scanner near the target building and performing 3D measurement, the return measurements get corrupted because of the reflectivity and transparency of the glass objects. The discrimination of reflections and indoor points from the outdoor environment is a challenging task. As the first step of removing these erroneous measurements, it is important to detect these glass planes framed in the building façade. In this study, we propose an unsupervised segmentation approach in combination with the threshold of Gaussian distribution to extract transparent and reflective glass planes from point clouds with intensity attributes. For practical validation, our approach is evaluated on two scan points of a building with many glass components.