Extracting individual bricks from a laser scan point cloud of an unorganized pile of bricks

Abstract

Bricks are the vital component of most masonry structures. Their maintenance is critical to the protection of masonry buildings. Terrestrial Light Detection and Ranging (TLidar) systems provide massive point cloud data in an accurate and fast way. TLidar enables us to sample and store the state of a brick surface in a practical way. This article aims to extract individual bricks from an unorganized pile of bricks sampled by a dense point cloud. The method automatically segments and models the individual bricks. The methodology is divided into five main steps: Filter needless points, brick boundary points removal, coarse segmentation using 3D component analysis, planar segmentation and grouping, and brick reconstruction. A novel voting scheme is used to segment the planar patches in an effective way. Brick reconstruction is based on the geometry of single brick and its corresponding nominal size (length, width and height). The number of bricks reconstructed is around 75%. An accuracy assessment is performed by comparing 3D coordinates of the reconstructed vertices to the manually picked vertices. The standard deviations of differences along x, y and z axes are 4.55 mm, 4.53 mm and 4.60 mm, respectively. The comparison results indicate that the accuracy of reconstruction based on the introduced methodology is high and reliable. The work presented in this paper provides a theoretical basis and reference for large scene applications in brick-like structures. Meanwhile, the high-accuracy brick reconstruction lays the foundation for further brick displacement estimation.