This paper is part of the 2017 3D Digital Documentation Summit.
Using UAVs to Rapidly Preserve Historic Structures in New Orleans, Louisiana
Alanna Moore: OK, so as the last presenter, I want to say thanks to everyone for a really great week. I’m self-taught in all of this, and I really, I have multiple questions I’ve had for years answered in the past two days, so it’s been really great for me.
So as Jason, said, my name’s Alanna Moore. I’m a master’s student at the University of New Orleans in the Urban Studies program, and I’m currently working at a culture resource management firm and also studying that in school. I’m trying to apply digital means of preservation to vulnerable historic areas.
So in this presentation, I’m going to talk about a 3D model that I made, using a commercial UAS drone of a shotgun house in the Bywater historic district of New Orleans, and I’m going to talk about, also, the broader implications of what this means for historic preservation, especially in Louisiana.
OK, so for this project, I’m taking a very different approach than all of you guys. I wasn’t concerned about accuracy. I more or less wanted to see what is the bare minimum of what I can do in the shortest amount of time with no money, because as I said, I’m self-taught, doing this all on my own. So with that, I was also trying to, essentially, establish a workflow of how this could be taught to people that are not photogrammetry, that are not drone pilots, that are only people interested in conservation.
So traditionally, I’ll breeze through a lot of this ’cause a bunch of it is redundant, of course. So traditionally, structures are documented using scale drawings and film photos. And those require a skilled operator, and typically a team going out and doing measurements in the field. So that takes a lot of time. It takes a lot of money. It takes a lot of effort, and typically, the products that are created are just stored in archives. They’re only actually used by the professionals that are creating them, and that are studying them.
Whereas, with photogrammetry, with 3D modeling, with laser scanning, that really opens a lot of doors for getting your data out there, as well as what you can do with your data. So for one, the speed of scanning and the speed of photogrammetry is what I think is one of the most important things, especially here in Louisiana because it expedites the amount of data that you can create. So with a small amount of time and very minimal skills and very minimal funding you can generate functional models, and these models can be used to enhance curation, interpretation, and dissemination of information about cultural heritage. You could also expand the audience of people who get to access your information.
First of all, as y’all know, 3D modeling can be done on any scale. This is an example of a worm trail pot, it was excavated at Storyville a few years ago and its been locked in a box in U&O’s Archeology lab, so I’ve made a really quick scan of it. Took me maybe twenty minutes to put on Sketch Fab, and now its been shared with the whole department. They’re all really interested in the possibilities of this. I’m also trying to get it printed eventually. So again, this can be used for bigger structures and it’s really great for remotely located structures. This is the Irish Bayou Castle that’s located, I wanna say, about twenty miles away. It’s close to Slidell, it’s on the shore of Lake Pontchartrain. It survived Katrina, it survived a couple different hurricanes, but it still is very difficult to access. It’s about an hour and a half out. It is also a residence so you have to deal with that too. And, this can also be used for preservationists.
Just an example of perseverance lodge 3, which is in Cypress Grove Cemetery, its been struck by lightning a couple different times and so I was contacted by a woman who was doing the repair work for it in order to actually have a drone go up there and see to examine the crack itself, and I said, “Why don’t I make a 3D model?”, so we could actually do the measurements of it. In pics 4D I drew the crack out and it says that it’s a meter or so. So again, like I said, breezing through all of this, UAS photogrammetry works in a pretty similar way as to restoral photogrammetry and long range photogrammetry. But, since the sensor is so much closer to the subject that you’re recording you have a lot higher resolution of imagery and of your data outputs. Prior the the commercialization of the multi rotor drum, photogrammetry really didn’t have a way to capture a 3D view in high detail of a structure because they were using airplanes, they were using helicopters, they were using cranes, things like that. It was just a much more dangerous, much more labor intensive process before drones were in the game.
There are two main types of drones, the multi rotor and the fix wing drone. I typically use multi rotors for 3D modeling because they have the vertical takeoff and landing aspect of them and they can hover, which means you have very very very precise control of exactly where you want the drone to be. However, the down side of that is that their flight time is a lot lower because the motors weight a lot, the cameras weigh a lot, and it’s a higher pay load and because of that, you get to fly less time. Compared to a fix wing drone that only has a single motor, but you have a lot less control over where it actually flying. It operates more like an actual traditional plane. So you need space to take it off, you need space for it to make turns, et cetera, et cetera. But it has a lot longer battery life, so multi rotors I believe are best for 3D modeling. Fixed wings can be used for 3D modeling but they’re typically more used for large scale mapping.
I was going to mention how no one else had talked about Pix 4D, but then all of a sudden, all four presentations before me did. So Pix 4D is the software I have chosen to use, it’s the premiere program for specifically UAS imagery. You can create 2D data products, like Orthomosaics and Digital Drain or service models, copper lines, and vegetation indexes, as well as 3D data such as Point Clouds, Textured mesh, and 3D point clouds. Or 3D PDS, excuse me. So, which makes Pix 4D exceptionally powerful to me, especially, is its mobile app, Pix 4D Capture. The software allows for the user to design flight plans, specifically for 3D model generation. So using this app, the operators need to draw a polygon on their phone of the area of coverage desired. Select the type of flight that you’re trying to do and input the altitude and the photo coverage you’re needing and press “Go”, and the whole process is completely automated for you. So the app sends weigh points to the drone telling it exactly where to go and exactly when to trigger the camera to make sure that your model is of good quality. And these are some photos of the app at work. This is the newest version, the original version only allowed you to do a grid or manual flight. The newest version gives you these other options of drawing a polygon, doing a double grid, or doing an orbital flight.
So with that benefit of use, there is the hefty price. With the full version of Pix 4D mapper, our perpetual license is almost nine grand, and educational license is about two for a single computer. And so, to kind of offset that, they give you some options for individual services where you can get just the thing you need. This is only within the past two months that they’ve released all of this. As I’ve said before, the technology is exponentially increasing and growing and developing. This is very new that they’ve differentiated the full desktop program for the separate little aspects of it. You can buy Pix 4D for BIM for about ten grand, and Pix 4D model, which is only modeling, no 2D information for five hundred dollars per year or forty dollars a month. For some reason, there’s no yearly for Pix 4D Model. And then with the mobile apps, you get five free uses of the grid flight planning and unlimited once you get the software purchase. Also, I’m sorry it’s doing this clicky thing, it’s a ghost of a prior presentation that I didn’t know was going on.
For this, what I’m going to talk about is a case study of a house, I call it the Royal Street House, because I didn’t want to say my old apartment. I chose this specifically because of its location, it’s the circular thing to the right, is the Royal Street House. I put a star where we are right now for those of you who are not here, to kind of orient yourselves to where we’re at. And then, a few other neighborhoods around New Orleans.
Okay, the reason that I chose this site, besides the fact that it is my old apartment I’ve lived in one form of the landlords house. I lived on one side, he lived on the other and I’ve been his tenant upwards of six years now so he was very excited about me doing this. I knew all the neighbors and was able to get in touch with them and get approval for all of that. But mostly I selected it because of its location, because as I mentioned before, I’m doing a lot of this with the coastal component in mind. So the house is located on the last block before the Mississippi River as well as by the Industrial Canal, commonly called the Industrial Harbor, or something of that sort.
It is obviously very vulnerable to environmental factors. The house was built in the early 1900’s and it’s constructed from recycled barge board. So as the boats were going down the Mississippi, they would deconstruct them and then build them into houses in this end of the neighborhood. It’s within a historic district, it’s a third of a mile from the two major waterways, and it’s one mile south of where the levy’s failed for Katrina. It’s on the opposite side of the canal, so this house actually was not damaged in the storm. However, it was kind of alarming to see how close it was and the devastation that happened there. Additionally, there’s a proposal to start dredging the Industrial Canal and widening it for more shipping, so that could in some way impact the structure as well as remodeling this old navy base right here, these three buildings. They’re talking about making that some kind of residential, Disney World, cruise thing. So that could again, affect the structure.
Here is the house in Google street view, and here is the house in Google street view 3D, which I included because this is obviously something that people are very interested in and are trying to work towards, but there’s not really a way at the moment to do photogrammetry and to do 3D modeling on a large scale. So a very basic methodology is you plan your flight, you press a button and fly it, and then you process your data. I mentioned previously that my goal with this was to just see the very basics of what I could get with the least amount of time, money, resources involved.
I was originally only going to do one flight, but I bought my new drone and I wanted to check it out. So the first flight I did, I just opened the Pix 4D Capture app. At the time you couldn’t do orbital flight, so I had to manually fly it and I just pressed “Go” with the settings that were on the app when you first open it. Unfortunately with that, originally with Pix 4D capture app, it would trigger the camera based on the distance the drone had traveled on the ground, so whenever I opened the app, it was one meter by one meter distance. Which doing an orbital flight, mean there were tons of photos. I took three hundred and twenty two, which is absolutely extraordinary. It took over seven hours of processing, and as you can see in the bottom left photo, each green line shows a photo in which that point was visible. So as we were talking yesterday, I can’t remember exactly who said it, you needed to have nine view points of an image in order for it to register, and clearly this one has a couple hundred compared to my second attempt, which the app has developed my drones a lot better.
Everything was just better conditions basically, so I was flying this one with a Mavick, and instead of doing the distance that the drone flies, now the capture app asks you the degrees that you need between each photo, so with that I did ten degrees and then I put only thirty six photos, so ten percent less than my original drone but still about the same amount of time processing, which is interesting. Other than the processing time, the only real differences between the models is the one with more photos was very noisy. However it was a little bit better coverage because flying manually, I was able to fly lower versus doing it with the app, which set me at my lowest height being forty meters.
So for both processing of this, typically with 3D models you use ground control points in order to increase your absolute accuracy. While they’re not necessary, they do almost always improve it. Typically a rule of thumb is the era of your model will be about two to three times the accuracy of the coordinates you’re using to geo rectify it. So if you’re coming in there with an archeykay system that’s getting subcentimeter accuracy, you can expect that you’re gonna get about a couple centimeters of accuracy for your model. Versus if you’re using Google Maps, which I did in this instance, because like I said, I was trying to do this no budget at all, that will give you an accuracy of feet to meters because the accuracy of the satellite imagery you’re using is feet to meters off.
Also, one thing that can affect that is actually marking the points in the photos. The image to the right shows the photo marking process so I used the corner of the stairs of the apartment as a ground control point, and unless you’re actually able to zoom in as close as possible and click on the precise pixel where you’re ground control point is, you’re gonna be a few inches off based on that as well. So in using Google Maps, first you actually go to the satellite imagery, zoom in, and click on the point that you’re using as a ground control point, you copy the coordinates and go to a website such as Daf Logic, which will give you the actual Z elevation. And again, all of this is interpolated so this is introducing a lot of call for error into the process, but like I said, I wasn’t really going for accuracy, I was looking for visual quality.
So with the flights that I did, I flew an orbital flight, as well as a grid flight, as well as took terrestrial photos. But to get just the quick and easy I only processed the orbital flights and did almost no editing on those at all. You may be able to see, there’s a telephone pole in the front right of the house. I cut that out in all of them and I drew a processing area, so you’re not looking at a full thing. But that is the only editing that I did on any of the models that I’m gonna show y’all.
So with the deliverables Pix 4D can give you, you can get Rastery imagery, elevation models and contours, not in this instance because I didn’t do the grid, although that is an ortho photo, just to show you. And then, you can do fly through videos, point clouds, textured meshes, poly lines, and shape files. So the point cloud is the primary form of 3D video for backend users. Each point within the cloud represents and XYZ coordinate in space and has an RBG value associated with it. And so in using this model, metric information can be generated from almost any point on the surface of the structure without requiring prior measurements. And this file type can be generated quickly, stored indefinitely in reference back to whenever measurements need to be compared. So I added this photo just to show that one point, it gives you the exact coordinates of it, the 3653, that’s in state plain feed I believe.
What’s kind of tricky about point clouds is areas that are obscured. So for this, that banana tree means the measurements of the roof are gonna be a little bit off. You can’t really measure the window beneath the banana tree because it’s obscured and it just adds extra points you can click to, as well as the front has a bit of an awning and a back porch. Because those areas are obscured and because I did not use any terrestrial photos, they’re kind of warped in the point cloud as well as in the model. So they’re not really valid for great measurements in those areas.
The same issue exists on the opposite side of the house because its got a fence within two feet of it. It’s a very thin alley way, so it kind of prevented what the drone could see near the bottom of the house as well as my ability to take full coverage photos, terrestrial photos. As we learned in this process, the quality of the model is really dependent on its surroundings. So the Royal Street House seemed ideal to me at first, kind of to prove a point about its location, et cetera, et cetera, but then I looked back at it and realized it’s not so great because of all the vegetation around it, because it is in a dense urban area. All of these things kind of lead to a draw back. However, I want to say that my end result of a one flight model is not the worst thing I’ve seen. Good enough I would say.
So this photo illustrates the excess noise with the original model, the one that used three hundred photos. Because there were so many photos, I think it was basically connecting the shadows, trying to generate key points when there really were no key points there. So as you can see along the roof, kind of like around the cars, things that were moving, it’s really bubbly and not pretty. So it’s not a perfect model, however for pure documentation sense, I want to say that this is an interesting thing that this could be done by literally anyone who has a cell phone, access to a drone, and can press a button.
So all of us are much more familiar with textured meshes. This is what is typically curated, this is the front end user kind of thing. So meshes are created by triangulating the spaces between the points and the photos that are input to the model are stitched together, warped, and projected on the model to give it a realistic sense. This information can be easily uploaded and shared online indefinitely, endlessly.
Some challenges that I faced were the urban environment, it obscured a lot of areas, there were a lot of loopholes to jump through, et cetera. The area’s obscured because of the house, I chose this one because you can really see beneath the awning, it very much just warps up and turns into a bubble because there are no points there, so the software tries to connect to the nearest point. Other issues that I faced was weather. I was kind of planning this, planning this, planning this, so I had to keep pushing it back, which was not a big issue for me at all, but would be really difficult for people that are doing this professionally. Additionally with why I was approaching this, was the idea to have a tool set as well as a skill set that could be easily taught to people and easily picked up. And I kind of realized that in doing it, it’s not quite as easy as I make it out to seem. You do need to have a very basic understanding of a drone, the laws associated with it, and what could go wrong. As well as how you do the processing and things like that. So that’s something that I think as the software gets more commercial, more accessible, more people are gonna start learning about it and be able to understand it better and be able to talk about it more.
Additionally, another thing I think is a challenge is the evolving Government regulation. It’s kind of a daunting thing for people that want to get into this but don’t want to sign up for a certification, sign up for something and then a few weeks later, have to change it all up again. So that’s something I think is interesting. And then, to bring all this home, I want to explain to y’all why this is something that I do care about, and why I’m passionate about it.
In Louisiana, we’re used to seeing these things called Red Maps, which every few years, The Coastal Protection and Restoration Authority comes out with these maps and it illustrates what land loss and what land gain Louisiana is expected to experience in the next few years. This is with no action at all, all that is red will completely disappear. It already is holding on by a thread, but it will be completely gone. So I added this because, the first star to the top is New Orleans, and the star towards the bottom is the town that I came from, Homa. I grew up there until I was 18 and looking at this map, that’s something that is realistically not going to exist in a very recent time. Florida is represented here, we have a lot of people who are from coastal communities and I’m sure can feel the uncertainty and the fear that comes with the space that I grew up in is not gonna be there whenever I’m an adult. It’s not gonna be there when I have children.
So that’s kind of what has motivated me so much to work towards both expediting the process of doing documentation for historical structures, cultural landscapes, all of it, as well as expediting the process of prospecting archeology to find where sites are. And all of these things are possible with 3D modeling and with drones. With remote sensing, with all this new technology coming about, it’s kind of getting people behind it and getting them to realize the fear that I feel of wanting to take all of it in so that future generations can see it before it is too late.
So just to kind of illustrate more on that point, since New Orleans is not quite a coastal city, this is a map from a report that came out last year from the NASA Jet Propulsion laboratory, about subsidence in New Orleans, and the star is the site of the Royal Street House that I modeled, which is according to this map is experiencing between ten to twenty millimeters of subsidence every year. Which is wild, that’s a crazy number and a crazy thing to see these maps and to know what is coming and to know that few people are doing much about it to preserve it. There’s a lot of work in preserving the land and coastal restoration, but there doesn’t seem to be very much focus on all the resources that exist in Louisiana and New Orleans specifically.
And so, to that here is a map of the historic districts that are managed by the city, by the Historic Districts Landmarks Commission. I added that to offset this, because this is a pretty bad map, this is all I could find. So if either of the historical preservation, the national register historic districts, within New Orleans. As you can tell it’s pretty hard to find a place in New Orleans that’s not in a national register district. Literally, the whole city is just teaming with history, teaming with structures, that ought to be and deserve to be documented and with the way that it is now, with traditional methods, there’s no way you can collect all that information in a time that we are allotted here in this state.
So to that, here are some other sites that I am interested in looking at for the sake of Louisiana disappearing, one of them being Fort Proctor which is well on its way. It’s a submerged, nineteenth century fort, and that site is also listed on the national register, so it would have a much bigger implication than modeling my old apartment. Fort McComb also is another nineteenth century fort that’s located in New Orleans East. I thought that one was interesting because it was used as the site for Car Cosa on That Tree Detective, season one, if any of y’all saw it.
And, Hope Cemetery, which is what I’m writing my thesis on. It’s the only below ground cemetery in the actual city of New Orleans, and it’s actually interesting because it’s not decorated in the above ground stone structures that most of us are used to seeing with New Orleans cemeteries. It’s a lot more handmade and, I hate the word authentic, but I want to say authentic New Orleans, as far as cemetery and burial art go. With that, I’m trying to use my drone to make an interactive map and to collect all the information, all the headstones, and have a tool that people can use to view and query it. Find their relatives, find famous people, anyone they’re interested in, as well as search it by the actual burial art style, which is something I think is cool. And I want to with that, incorporate 3D models, and it’s a long shot, connecting with families and asking their permission, et cetera, but I was hoping to find interesting artifacts at the grave sites, make really quick 3D models with them in the field and then be able to incorporate that into this, but that’s all future work. Things I’m thinking of to basically document before they’re gone. That’s what is motivating me to be here and do this now.
This is a video of Fort Proctor, just to kind of put into perspective. I’ve been told by a lot of people that you don’t know what’s going on, you can’t feel the crisis in coastal Louisiana until you actually fly over the coast and you see. If you look at this on Google Maps, it shows it as solid land, and whenever you see it in this, it’s quite clear that it’s not, it’s disappearing. This is from an airplane, my professor took this and sent it to me of Fort Proctor, just to kind of show you guys. And thank you, that’s all.
In recent years, unmanned aerial vehicles, or UAVs, have become increasingly accessible to the public. Due to this expanded accessibility, users continue to develop new applications for low altitude aerial photography and photogrammetry. Photogrammetry is the science of generating three dimensional data using properly aligned photographs to triangulate measurements of a landscape or structure. With the incorporation of GPS data, these 3D maps and models can achieve about one inch relative and absolute accuracy, meaning very highly accurate geolocation as well as scale information. While most people are applying this technology in engineering and architecture by modeling current structures so that they can be added to using Computer Aided Design programs, we believe that it can also be used as a preservation technique for individuals or municipalities with limited cultural resource management budgets.
The gulf coast is one of the nation’s most culturally rich landscapes, with New Orleans at its center. New Orleans is renowned for its unique blending of French, Spanish, and Caribbean architecture styles and brightly colored buildings. With the oncoming threat that climate change poses upon this port city, the documentation of these structures becomes critically important so that future generations can continue to maintain and repair the historic structures. Digital 3D models allow people to virtually experience and recreate the landscape of the city. This project aims to present techniques that ensure that the historic character and unique landscapes of cities can be documented and made available to future generations.
Our presentation will discuss the methodology and application of capturing and modeling city blocks within a Historic district of New Orleans, Louisiana. We will be presenting a technique which demonstrates a quick, affordable, and accurate means of historical documentation, and in our presentation we will discuss potential applications of 3D data as an educational tool.
? Use of drones/ UAVs for collecting imagery
? Climate change
? Community Engagement
? Three dimensional models
Alahna Moore is currently a graduate student in the department of Urban Studies at the University of New Orleans. She is certified in ArcGIS through ESRI, and remote sensing for land management through NASA ARSET. Alahna has ample experience using GIS to conduct spatial analysis using satellite imagery through her work with NASA DEVELOP, and has spent the past year generating terrain data for land surveyors using unmanned aerial vehicles and RTK GPS systems. In academia, her work has focused upon incorporating GIS, UAVs, remote sensing and photogrammetry to aid in archaeological excavations and digital modeling of historic structures. Alahna’s primary research goal is to expedite the processes of documenting and monitoring coastal landscapes that are increasingly threatened by the effects of climate change.