Applied photogrammetry for 3D modeling, quantity surveying, and augmented reality in construction

Abstract

The surveying technique of close-range photogrammetry is based on an analytical representation of the image forming mechanism of photography and extracts spatial information through computation on photos. This research reviews the fundamentals of close-range photogrammetry and applies close-range photogrammetry to (1) model 3D construction graphics, (2) measure the geometric dimensions of building products, and (3) augment site photos with 3D graphics of underground facilities. This research firstly establishes a 3D modeling method based on the mechanism of photogrammetry to ease the effort in providing model ingredients for visualizing main processes and major products in construction operations. The current way of 3D modeling, which relies on the use of computer-aided design (CAD) or proprietary software for virtual reality (VR) development, not only requires object design drafts or geometry specifications to be prepared beforehand, but also has a tedious, time-consuming procedure. The proposed method utilizes a digital camera to capture site images and analytically processes the image data into a 3D model of an object, by which the effort of modeling is largely alleviated. A precast facade is modeled under practical constraints to verify the method. The method is useful particularly when the design drafts or geometry specifications of the modeling objects are unavailable or the modeling objects are inaccessible for direct measurement due to safety concerns. The photo-based 3D as-built modeling method provides an alternative to taking geometric measurements on building products. In contrast with the conventional measurement tape, applying photogrammetry is cost-effective and safe as the measurements are conducted on the 3D model of an object resulting from photos, instead of on the object itself. Thus, the proposed method is conducive to measuring the building elements situated in hazardous areas and quickly checking the dimensions of precast units on site for quality control purposes. The measurement errors of the photo-based method are attributed to (1) the systematic error due to camera lens distortions and (2) the random error due to human factors. Seventy-nine paired measurements (length, width, and height) are sampled on twelve structures and facilities by applying the photo-based method and measurement tape respectively. The 95% limits of agreement are established on the sample data to statistically characterize the accuracy level of the photo-based method against tape, resulting in [-15.30 mm, 11.39 mm]. Through weighing the accuracy level against the accuracy level desirable in a particular application, the engineer makes the final decision on the applicability of the photo-based method. In addition, this research adapts the analytical algorithms of photogrammetry into a computationally simple yet practical method for incorporating computer-generated, three-dimensional (3D) as-built graphics of invisible underground infrastructure into site photos, resulting in a richer and more integral view of the site situation. Previous photo-augmenting methods require both the camera's position and orientation to be determined on site. The new method simplifies the superimposition of a view taken from a 3D virtual model onto an actual photo by analytically fixing the camera's orientation with the coordinates of two reference points only, namely, the camera station position and the object focus position. The advantage of the proposed method over current photo-augmenting techniques lies largely in alleviating the effort required for camera positioning in the field. By setting a virtual camera in the 3D modeling environment in the same way as the site photo is actually taken, a virtual view of the underground scene is produced, which is then analytically merged with the site photo by coinciding the real and virtual coordinate axes in the two-dimensional (2D) image space. The new approach has been applied to superimpose as-built models of infrastructure onto site photos in order to facilitate quality check of bored pile excavation and progress visualization of micro-tunneling construction. The level of accuracy for augmented photo registration depends on the specific implementation settings on site including the surveying instrument applied and the camera used. As for the bored pile excavation and micro-tunneling cases conducted in this research, given the exact coordinates of the camera station position and the object focus position, the augmented photo registration can achieve an accuracy level of 0.07 mm on the camera image plane.