Recognition of temporary vertical objects in large point clouds of construction sites

Abstract

Although adherence to the project schedule is the most critical performance metric among project owners, still 53% of typical construction projects exhibit schedule delays. While construction progress monitoring is key to allow effective project management, it is still a largely manual, error-prone and inefficient process. To contribute to more efficient construction progress monitoring, this research proposes a method to detect automatically the most common temporary object classes in large-scale laser scanner point clouds of construction sites. Finding the position of these objects in the point cloud can help determine the current state of construction progress and verify compliance with safety regulations. The proposed workflow includes a combination of several techniques: image processing over vertical projections of point clouds, finding patterns in three-dimensional (3D) detected contours and performing checks over vertical cross-sections with deep learning methods. After applying and testing the method on three real-world point clouds and testing with three object categories (cranes, scaffolds and formwork), the results reveal that the authors' technique achieves rates above 88% for precision and recall and outstanding computational performance. These metrics demonstrate the capability of the method to support automatic 3D object detection in point clouds of construction sites.