ASSESSMENT OF OBLIQUE PHOTOGRAPHY IN ENHANCING SUAS-BASED STRUCTURE FROM MOTION PHOTOGRAMMETRY-DERIVED DEMS

As small Unmanned Aerial System-based (sUAS) data collection becomes increasingly prevalent, it is imperative to consider methods of enhancing such collection by improving accuracy and efficiency. It has been hypothesized that acquiring a combination of oblique and orthogonal photographs of the earth’s surface may yield higher quality digital elevation model (DEM) and orthophotographs. Using a DJI Phantom 4 Professional sUAS, sub-decimeter accuracy aerial photographs of Dry Creek in south-central Missouri were captured at camera angles of 50⁰, 70⁰, and 90⁰ (with 90⁰ being orthogonal to the land surface). Both north-south and east-west-oriented flight missions were used for each camera angle, producing six unique sets of photos varying in angle and orientation.

These photo sets were then processed using structure from motion photogrammetry (SfM) algorithms in Agisoft Photoscan (v. 1.4) software. To aid in the photographic alignment and test elevation accuracy, 11 ground control points were placed within the research area which were located with RTK GPS to an accuracy of <10cm.

In Photoscan, DEM resolution and point density is controlled by processing quality. Thus, the inclusion of oblique imagery has the potential to improve the accuracy of the resulting DEMs by increasing the quality of the resulting point cloud used to construct the DEMs. The inclusion of oblique photos in the SfM processing significantly increased the time and computational effort required to create the DEMs. However, no correlation was found between the number of images or their camera angle and the horizontal and vertical accuracy of the DEMs.

The inclusion of oblique photos in the Dry Creek data did not improve DEM accuracy. Other investigators have reported that oblique images improve DEM accuracy, particularly in study areas with prominent vertical features such as buildings, forested areas, and cliffs. Perhaps, the lack of such features within the Dry Creek area limited the utility of the oblique imagery in relation to our DEM accuracy. Our findings suggest that an orthogonal image dataset processed at the “ultrahigh quality” setting provides the best “bang for the buck” in terms of DEM accuracy. This conclusion is an important insight for planning and designing sUAS-based SfM missions.