This presentation is part of the 2017 3D Digital Documentation Summit.
A Bird’s Eye View: Photogrammetric Documentation using Drones
Speaker 1: Hello. As he said, I’m a PhD student. You guys are a tough act to follow. In fact, Kristen, you probably don’t realize it, but you’re my mentor. We had contact about a year ago, and I was just getting into drones then. I had seen an APT demonstration in Boston, and it was her company that did that demonstration. They inspired me to use my drone that I had been using for photography in another way, so thank you for that.
You obviously have come a lot farther than I have, but I did do this project with a class. We did a documentation project of a church that happens to be across the street from our university, and we used an awful lot of documentation techniques on it. It has a very high roof, and I thought it was a good opportunity to try my drone skills, which are in the beginning, but I think I got a pretty good result. Those of you who are interested in getting into drones, I encourage it highly, and hopefully you’ll see how far I’ve come in a little time and you’ll be encouraged.
So photogrammetry, everyone here, I’m sure, knows the history and what photogrammetry is. I’ve put this in here just for a little bit of a human element because I don’t think a lot of people realize that this Albrecht Meydenbauer was actually the developer of photogrammetry in Germany in the late 19th century. He was a really interesting guy.
He actually developed photogrammetry for buildings. We use it for a lot of different things, mapping, and we use it for statues and paintings, which we saw some of yesterday. But it was actually developed for buildings, and it makes me very happy that we all are now using it for cultural heritage.
He developed that institute, and they had a 20,000 photo plate archive. You can see here that when Berlin was bombed in World War II by the allied forces, it destroyed this French Cathedral. They used Meydenbauer’s plate to do the reconstruction in the 1970’s. Those plates that did survive the war got put to good use.
Unmanned aerial systems, we call them drones, it’s interesting they were developed for the military, but they’ve come a long, long way in commercial development. The consumer models have gotten very, very good. It says here that they’ve logged three million flight hours in Afghanistan and Iraq. I didn’t realize that myself as a civilian, but what I thought was really interesting is this forecast of the market.
It was done by Radiant Insights. They did this forecast, I think, in 2014 or 2015, and it’s staggering how popular drones are. In fact, I think it was the number one Christmas present last year, so the consumer models, they’re getting better and better, and I have a consumer model myself. I’ve seen the other ones, and I drool over them frankly, but I can’t afford them either.
In the commercial world, drones are doing a lot of things. You can see a listing of things here. Some of the things that may not be listed here: in the energy industry they’re using them for oil and gas mapping. They’re also using for high powered tension utility line inspections. In the shipping world, they’re using them for yard management, and there’s actually proprietary software that they can go and manager their container system, find them, and that sort of thing. In the warehousing industry, they’re using them for inventory tracking. In the future, they’re gonna be using them for commercial pipeline maintenance and also border patrol, so maybe we won’t be getting a big wall. Maybe we’ll have drones flying around.
This is the pilot study that we did. Our class did this documentation project, and we started out using traditional HABS level, hand drawing, field sketching techniques. That was the focus of the class actually, but my fellow graduate students … Soojin you heard yesterday. We took the opportunity to use some digital technologies on this project, and he was very good at doing the laser scanning, and I decided to try out my drone skills, my newly acquired drone skills. And I also did the heritage modeling for this project.
My focus of my dissertation is the adaptation of building information modeling for heritage. However, I wanted to go out on a limb and do this special little side project for this building. You can see the very tall roof. It didn’t allow us access, and even if we could’ve gotten up there on a crane or a lift, those tiles are custom 1961 mid-century modern tiles, and we were very afraid that we might hurt them.
You can see that, in some places, maybe not in the photograph, but in real life, you can see that they’re still kind of clipped on because they don’t wanna replace any of them and they don’t want them to fall down, of course. Where they’re loose, they just clip them up with these metal clips.
My goal was to create a point cloud and see if I could merge that point cloud with our terrestrial laser scanning that Soojin was doing. You all saw some of his work yesterday on this church. I was concerned that the point cloud, the data might not be robust enough to be able to be merged well enough.
So that was my goal, to see if it was. I used my Phantom 4. I was inspired, like I said, by Vertical Access. They demonstrated this model in Boston, and I was very impressed with it. I didn’t know about the rolling shutter when I bought it. They shortly thereafter came out with the Phantom 4 Pro, which has eliminated that rolling shutter problem, so I had to figure out a work around.
I did by using Pix4DCapture. I used Pix4DCapture to take the images, and I found out that if you use the Pix4DCapture and you use the pre-program flights, you can choose to stop the drone to take the image. The rolling shutter problem really comes about when the drone itself is moving during the image capture or if the subject is moving, which we have a building, so that’s not a problem. So I just made it stop at each way point to take the image. It made the flight a little bit longer, but luckily, the building, it wasn’t so large that I ran out of battery during the flight.
It has very good specs. Someone asked me yesterday what the megapixel of the camera image was. 12.4. I said I would find out for you. It has a 2/3rds of an inch [inaudible 00:08:10] sensor. I’m used to using a full frame DSLR, so this is very small for me, but the images were quite remarkable, I thought. The challenge … One of the challenges that I found with it, though, is the limited pitch of the camera. You’ll see one of the results that I got has something to do with that.
Yes, the FAA rules changed. Someone else mentioned this, and you had a different date so I could be wrong with my date, but it was last year, late in the summer. Everyone then had to get remote pilot licenses. It’s not a very difficult test. I’m not very math oriented myself, but it was easy enough for me to pass, so I encourage you to go and take it. But you only have to have it if you’re going to be doing commercial drone flying. If you have to ask if you’re doing commercial, you probably are.
Hobbyists know if they’re a hobbyist, but unfortunately for me, in my situation, I’m not doing it commercially, but I am doing it for research. There is a small, little rule that says, if you do it for academic research, you have to have a remote pilot license, so I got one. If you’re doing it just for educational training; however, you don’t have to have one. There’s just a small distinction in the rule about that.
The apps that I use are the Before You Fly app, and I also use UAV forecast. I find them to be very easy to use. I have Air Map, but it’s a little bit more difficult I’ve found, so I don’t use it as much.
This is just a sampling of the flights that I made over the church. There were some constraints flying over this church. It is on a very busy highway … Not highway. I’m sorry. A very busy road, traffic corridor. It’s this road right here. It’s right across the street from our university, and there’s a lot of foot traffic, a lot of students walking around. There happens to be a bus stop right there in front of that church. When you’re trying to avoid flying over people, I had to have a spotter who came out and made sure that people weren’t there. I had to wait for the buses to come and go and unload, that sort of thing.
Also the weather. In Florida, we have afternoon thunderstorms, so I had to do some planning. The first speaker was right. It’s all about planning. These are just some of the flight paths, you can see, that I took.
I found the flight path app, Pix4DCapture, to be very, very helpful. I tried a few manual. I commend you if you can do that. I tried a few, and it’s very, very difficult to work the camera and know what you’re doing as far as exposure and distance and hold it steady and press the shutter and fly. To me it was very difficult, and I found that the pre-programmed flight patterns were a lifesaver for me.
Of all of the paths that I took, I chose, after I processed them in AutoDesk Recap, I didn’t use Pix4D. I chose AutoDesk Recap because they have an agreement with our university, and they give it to us, which is nice. Pix4D doesn’t have that same agreement with us. With Recap, I processed all of the flights, and I chose the one that I felt had the most data to it. It happened to be this circular flight path that I took at five degree intervals.
This is the final point cloud that was processed through Recap. You can see, I was only really concerned with this top portion, the roof portion, because we had the laser scanner. Since I was trying to augment our data with the laser scanner, I just was concentrating mainly on the roof.
This is the laser scanner data, and if you pay close attention to this portion here and right here, you’ll see how the drone point cloud augments the laser scanner data. I’ll just flip back and forth a couple of times ’cause it’s a little easy to miss.
So you can see it cleaned up that roof eve rather well. Soojin was nice enough to help me to analyze this data, and he informed me that the accuracy was not as good as the laser scanner, of course, but it was satisfactory. One of the things that helped a lot in the meshing of the data points was that the drone, the Phantom 4, has a GPS. So they were all geo-referenced. The point clouds came out geo-referenced, so they laid right over that laser scanner point cloud.
However, this was another bit of an issue that I had. You can see the point cloud section on the left, or your right, that is the drone. And the one that’s got the profile, that’s the laser scanner. So this is a section of the clay tiles, and you can see that the laser scanner in shooting up, got that really nice profile for those clay tiles. However, the drone, shooting down from the sky, it was very flat. You can also see a density difference. It didn’t really affect the fact that we were able to use some of the data to fill out the laser scanner point cloud, but it is a noticeable thing when you cut a section through the two point clouds.
As I said, I feel like it was pretty successful, our project was pretty successful in this very rudimentary pilot project that I did to practice my drone skills. I did learn a lot of lessons, and I would do some things differently next time, which, may not be a next time, but we’ll see, because I’m really more into my research with BIM, but it was very interesting, and I would add more flights. More flights and combine the point clouds to make the data more robust. Of course, the professionals, they do this sort of thing. I would also do a plan view, which I did not do.
My conclusions are that the point cloud data is robust enough for augmenting it with our laser scanner data, and I think that it also reinforces traditional methods. We’ve been having a lot of conversations about traditional methods versus digital methods, and I really do feel that they reinforce each other. They compliment each other.
When we did this documentation project for this church, we did traditional methods, and we were learning traditional methods, but we were also using digital methods. It’s really interesting how you switch back and forth so easily. You think you’re learning a new language, but you get to that point where you don’t realize if you’re using the traditional or the digital because it’s just so fluid a movement back and forth, especially for the younger students.
They grew up digital, so they are very familiar with the digital. Older students like myself, I tend to like traditional, but the digital excites me, and I do use it. I used the point clouds when I was doing the BIM model, the heritage model, and I used the original drawings as well. You have a mixture of the traditional and the digital. Thank you.
Technology is rapidly changing the way historic preservation specialists document and assess historical and cultural resources. Efficient methods for data collection are at the forefront of many preservation conferences and workshops. Much focus has been placed on 3D imaging technologies like laser scanning and photogrammetry as they are becoming more widely used for recording heritage sites. Developed in the late nineteenth century, photogrammetry is the science of acquiring measurements from photographs. Like many technologies of the period, photogrammetry developed rapidly during World War II and played a significant role in post-war restoration. The wider-scale acceptance and utilization of point cloud producing technologies, like photogrammetry, has led many preservationists to seek innovative solutions for streamlining the data collection process. This paper presents the workflow process and outcomes of a pilot study that utilized an unmanned aerial vehicle (UAV), commonly referred to as a drone , to conduct photogrammetry of a mid-century modern church. The tall, pointed roofline and character-defining copper spires of the building posed a challenge for access, making it a good candidate for the use of photogrammetry via drone. Among the benefits, drone-based photogrammetry provides preservationists with opportunities to acquire geo-located data of inaccessible areas of heritage resources, more cost effectively than aerial photography and, more safely than traditional harness and rope rappelling methods.
However, there are limitations due to recent Federal Aviation Administration licensing requirements, the operation of drones does require specialized training and a Remote Pilot Airman Certificate. There are often many uncontrolled site conditions that need to be addressed. One limitation of the drone used for this study is the rolling shutter mechanism of the built-in camera. A rolling shutter is a capture method which exposes the camera sensor line by line from top to bottom instead of exposing, like the traditional DSLR camera, the entire sensor simultaneously. Rolling shutter cameras sometimes produce artifacts on the image and extra care is needed in the acquisition process. This paper explores these and other outcomes in more detail and offers a set of considerations when recording heritage resources with drone-based photogrammetry. Demand for drone technology has increased the quality of drone cameras and flight software while decreasing price points. The rolling shutter limitation of the Phantom 4 has already been mitigated with the release of the next generation of this drone. Future improvements in drone technology promise increased flight control and automation which will lead to increased maneuverability and quicker data acquisition. These improvements and lower prices will increase the value of drones for historic preservation.
Marla Holden is a second-year PhD student at the University of Florida at the Center for Advanced Construction Information Modeling, where her research interests include the application of technologies to cultural heritage resources such as historic building information modeling (HBIM) and unmanned aerial vehicle (UAV) photogrammetry. Marla returned to graduate school after working in the construction industry for 20 years as a Preconstruction Executive specializing in High-End Resort Hotel Development. In addition to her studies and research, Marla has begun applying her industry knowledge and experience as a Graduate Teaching Assistant in the classroom. Upon completion of her degree, Marla’s goal is to obtain a posting as a lecturer professor and help educate the constructors and preservationists of the future.
Morris (Marty) Hylton III is Director of the Historic Preservation Program and Preservation Institute Nantucket at the University of Florida’s College of Design, Construction and Planning. His research addresses multifaceted strategies for documenting and advocating the preservation of endangered heritage sites, particularly cultural resources associated with modernism and the Recent Past and historic places and communities endangered by sea level rise and other threats. Marty also created and manages the Envision Heritage initiative to explore how new and emerging technologies like laser scanning can be used to document historic sites and cultural resources.