Photogrammetry: 3D Modeling Landscapes and Structures
Photogrammetry is rapidly becoming more and more versatile and accessible for professionals and non-professionals in the development of 3D models from 2D images. The process is particularly straightforward and does not require any special equipment. Agisoft Photoscan is software that is becoming more popularly utilized by archaeologists for site surveys due to its efficiency, cost effectiveness, and versatility.
Here I will discuss the challenges and successes I faced employing photogrammetry for the 3D modeling of large scale structures set within an even larger landscape. For this project I chose the northern entrance into Yellowstone National Park where a large sign sits at the base of Gardiner Mountain and greets visitors as they drive into the park. The iconic structure of the northern entrance is the famous Roosevelt Archway built in 1903, which was at the heart of a major $22 million facelift for the park’s 100th anniversary. However, a more recent addition to the north entrance is a large Yellowstone National Park sign that is located in a courtyard of sorts that backs up to the park’s landscape-making it a prime spot for tourists to stop and take photos. There is an interesting parallel between the size of the sign structure and the size of the Yellowstone landscape itself; both create a feeling of smallness and grandeur for any visitor that stands beside it.
The size of the structure was certainly intimidating when thinking about how I was going to capture its entirety for 3D modeling. I used a Cannon SLR camera to take several overlapping photos and began shooting on the side closest to the NPS arrowhead logo moving counter-clockwise around the structure. However, due to there being a fence, it was difficult to take consistent photos while trying to photograph the backside. The height of the structure also proved to be an obstacle to get full coverage of the sign. I attempted to use some of the surrounding rocks to get different perspectives and potentially the top of the sign.
After obtaining all the necessary photos, the next step of creating my 3D model was converting and importing all the photos into Photoscan. For Photoscan to build a 3D model the photos must first be aligned through common points, then individual data points are generated to form a dense cloud, a mesh is then wrapped around the dense cloud, which creates the surface of the model, and finally a color surface is laid on top of the mesh through the application of texture. My first attempt to align all 75 photos was unsuccessful. After revisiting my photos, I discovered that I had zoomed in on the structure in several of the photos, which causes distortion in the 3D model. In order to fix the alignment, I decided to go through and delete all the zoomed photos. Deleting the unusable photos allowed for the proper alignment of photos and I was able to move on to generating the dense cloud. Once the alignment issues were solved, I didn’t run into any other problems as far as creating the dense cloud, mesh, and texture. However, after all layers were complete I noticed that the back of the sign was not completely captured. This was most likely caused by not being able to fully circle around the sign due to the fence. Pieces of the fence and the full concrete pavement were also captured alongside the model. This however, I did not necessarily consider negative because it showcased portions of the landscape surrounding the structure.
Overall, I considered this project to be successful and I was able to troubleshoot the Photoscan software for the first time. This allowed me to expand my skill set and knowledge of 3D technology and its uses within the field of archaeology and heritage education.
