Assessing the quantitative and qualitative differences between 3D point clouds produced from lidar and structure from motion

Abstract

[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI--COLUMBIA AT REQUEST OF AUTHOR.] The purpose of this study is to compare the quality and accuracy of point clouds and digital elevation products derived from Unmanned Aerial System (UAS) acquired Structure from Motion (SfM) as compared to those obtained from lidar data. If the accuracy of UAS SfM proves to be on the same order as lidar, the technique could be used more often in lieu of or in addition to lidar. Three UAS SfM models were produced to compare with existing lidar. Two of the models were culturally important anthropogenic structures and the third was part of a prairie landscape. For each model a UAS was used to acquire photos. SfM software was used create 3D point clouds from the photos. The point clouds were then qualitatively and quantitatively compared to the lidar from the same areas. It was found that SfM point clouds could be produced that were statistically equivalent to lidar. In addition, it was found that the point count of lidar could be densified by up to two orders of magnitude while adding RGB values. SfM was also found to be able to model the undersides of structures which airborne lidar is incapable.