Document Type

Conference Proceeding

Publication Date

11-2020

Publisher

American Society of Civil Engineers (ASCE)

Source Publication

Construction Research Congress

Source ISSN

9780784482865

Original Item ID

DOI: 10.1061/9780784482865.032

Abstract

Image-based 3D-reconstruction techniques, such as drone photogrammetry, have been adopted to various construction operations. The challenge is determining the construction site elevation, which is the vertical distance from the ground to the camera, in real-time. This paper presents the research results of using two frame drone-based ortho-images to estimate the elevation of a construction site. This idea is derived from the stereo vision model for measuring distance of indoor scenes. The spatial resolution of the stereo vision is a positive correlation with its baseline, the distance between two cameras. However, a large baseline stereo cameras system is impossible for a drone to carry. Therefore, a modified stereo vision method is proposed to use a drone’s camera to capture low and high ortho-image pairs, which enlarges the baseline, and makes the spatial resolution adjustable for every single construction site. For determining the elevation of each pixel in the image pair, the researchers conducted drone flight planning, image pairs capturing, image transforming, image pixel matching, elevation recovering, and modeling. A field experiment was performed for evaluating the accuracy of the proposed method. The success of this research will advance the efficiency of construction operations that heavily depend on elevation information such as the excavation operations and facility layout.

Comments

Accepted version. Published as part of the proceedings of Construction Research Congress, 2020: 296-305. DOI. © 2020 American Society of Civil Engineers. Used with permission.

Download
Bai_14762acc (2).docx (4926 kB)
ADA Accessible Version

Share

COinS