Generation of Dense 3D Point Clouds Using a Small Quadcopter

Abstract

The disruptive technology of small and light-weight Unmanned Vehicle Systems (UVS) is changing geo matics applications and creating new and innovative opportunities for measuring and modeling outdoor and indoor environments. UVS’ compliment, and in many cases replace, total stations and laser scanners, and operate as flexible mobile survey systems. A small aerial quadcopter has been used to capture oblique highly overlapped video images. The video images captured from the onboard camera were reconstructed based on an incremental structure from motion bundle adjustment approach. Dense multi-image matching was used to generate a 3D point cloud representing the surface of the mapped area from the oblique video images, while the position and orientation of the image frames were also estimated in the process. The pho togrammetrically-generated dense digital surface model was co-registered and compared to an existing 3D building model. The paper presents the methodology and the obtained results and accuracies.