Activities #10 and 11: Aerial Mapping with a UAS

Introduction

Thanks to the ease in which people and organizations can share information and data over the Internet, aerial imagery for almost anywhere can be obtained through many different organizations or government entities. These images, however, may not always be the best option. Perhaps you want imagery for a very specific location or at a very specific time. Depending on what your project is, georeferenced images found online may not work. Using a UAS to collect your own images tailored specifically to your project may be the best way to go.

So far in this class, we have talked about and been introduced to UASs several times. This activity, which took place over two class periods, gave hands on experience using two different types of UAS: a multicopter and a balloon. In both cases, cameras attached to the UASs collected images of the surrounding area. It was up to us to mosaic these images together and georeference them so that the images would be correlated with their real-world locations and would be able to be used in programs like ArcMap.

Study Area

Figure 1 - Location of Eau Claire Sports Center

For both of these classes, our class returned to the Eau Claire Sports Center (Figures 1 and 2). This is the most convenient place to test UAS methods since it has several soccer fields, meaning there are few trees, electrical lines, or other obstacles that may cause issues with aerial devices. Plus, since we were doing these activites in early April, the soccer fields were not yet being used. We had a large, clear field all to ourselves.

Figure 2 - Eau Claire Sports Center. Notice how wide open the
area is. This is a great place to practice collecting aerial
images because there are very few obstacles.

Balloon

Methods


The balloon was inflated with helium in the parking lot of the sports center (Figure 3). Once filled, the picavet is attached to the line for the cameras to attach to. A picavet is a suspension-based rig that allows the cameras to stay relatively level with the ground, even if the balloon or kite is not (Figures 5 and 6). We used two different 12 megapixel cameras; one that produced images that were already georeferenced and one that did not. The cameras were set to collect a maximum of 300 images, capturing every 5 seconds.

Figure 3 - Inflating the balloon.

Figure 4 - Cody likes balloons.

Figure 5 - Attaching the Picavet.

Figure 6 - Picavet in action.

Figure 7 - Walking the soccer fields with the balloon in tow,
collecting aerial images.

The balloon was sent up to about 500 feet. At this height, the photos taken by the cameras would include a large area, and would overlap well when we mosaiced them later. As a class, we walked around the soccer fields taking turns holding the balloon (Figure 7). For the most part, we did not walk in any specific pattern, but we made sure that we covered as much area as we could. Only at one or two spots did we have to be careful of trees or power lines. For the most part, however, we didn't encounter any issues.

When we finished, the cameras were removed and the balloon brought down. Professor Hupy decided the most efficient thing to do with the balloon was to make it explode all over himself and the inside of a students car.

With the images taken, it was time to mosaic and georeference those images so the area as a whole could be analysed and the image could be used in programs like ArcMap. One of the students in our class, Drew Briski, was a huge help in this process. He was the first one to play around with the data and get it mosaiced together, and he taught the rest of the class what he learned. Here are the steps I followed for the next part of this post.

Figure 8 - All the tools needed to mosaic the images are
found in this menu, Workflow

There are several programs that be used to mosaic these images. I chose to use PhotoScan because Drew had such good directions for this program, and because images do not need to be georeferenced to mosaic them together. This can be done after in a different program. First I tried mosaicing images from the camera that georeferenced the images (Canon SX260), thinking that I would be able to skip that step then after I had my final TIFF. I only used about 25 images for this, since using all of the images would have taken several hours to process. Figure 8 shows the workflow menu in PhotoScan. Almost all of the tools I needed to use were in this drop-down menu. Add Photos, at the very top, is where you select the photos to use. Once those are added, I clicked Align Photos. This creates a point cloud of the images. Once that process was completed, Build Mesh was selected, which builds a TIN (triangulated irregular network) from the point cloud, followed by Build Texture. Once this is finished (Figure 9), the image can be exported as a TIFF file by going to File, Export Orthophoto. Since these images were already georeferenced, I brought the TIFF into ArcMap and ended there.

Figure 9 - Completed mosaic from the SX260 images. 

Figure 10 - 110 images from the Canon Elf mosaiced together.

After seeing the results, I decided to try the other set of photos from the Canon Elf. I used the same process in PhotoScan as I did with the previous set of images. This time, however, I used 110 photos to see how much better quality the results would be. Figure 10 shows the mosaiced image before I exported it as a TIFF.

Figure 11 - Georeferencing toolbar.

Since these images were not georeferenced when taken, the mosaiced image had no spatial reference. To georeference it, I brought it into ArcMap, added the georeferencing toolbar (Figure 11), and added the Imagery basemap so I would have a spatial reference to compare my TIFF to. On the georeferencing toolbar, I clicked the icon with the magnifying glass so that my unreferenced TIFF was opened in a different window. To add control points, the icon on the left (Figure 11) was selected.

Figure 12 - Adding control points in ArcMap.

Adding control points gives spatial reference to parts of the unreferenced image. For example, the first control point could be the corner of a house. I would click on that corner first on the unreferenced image, then on the same corner on the basemap. This tells the image that that is where that corner exists in real life. building corners, sidewalk corners, or other 90° angles always make great control points because they can be easily distinguished. The more control points you use, the more accurate the final image will be. Figure 12 shows an example of this. I used about 30 control points for this image.

Results

Figure 13 - Close up of the mosaiced image from the Canon Elf.

Figure 14 - Finished mosaic from the Canon SX260. Only
25 images were used for this one. This camera included
georeferencing for the images, but notice how off the image
is from the basemap below it.

Figure 15 - Finished mosaic from the Canon Elf. 110 photos
were used for this mosaic. This image I georeferenced myself
using control points in ArcMap.  Notice that this one is far
more accurate than the first try.

Discussion

The second mosaic using the Canon Elf turned out much better. This could simply be because I used 110 images instead of 25. That aside, the images that were already georeferenced were not very accurate. The camera was able to tell that the images were taken at the Eau Claire Sports Center, but that was about as close as it got. In the end, manually referencing the images proved to be much more accurate. Also, the images from the Elf were brighter and had much better contrast, making the finished product much better looking. Dealing with this one took much longer, however. It took a lot longer to mosaic, then I had to georeference it myself. When I brought the image into ArcMap, the image was mirrored, which made it difficult to add control points at first. But the image was correcting itself as I added more and more control points, so eventually looked correct.

Multicopter

Methods

Figure 16 - Joe's Y6 multicopter shown with remote control
and the laptop running Mission Planner.

While the end product (aerial photos of a specific area) is basically the same as using a balloon, the methods are quite different when using a rotary UAS. Since a balloon is fairly cheap, and can be operated by a single person, there isn't as much preparation that needs to go into it. A rotocopter, however, is a very expensive piece of equipment and, while it can be operated by a single person, should have a team of two or three people in control. For this assignment, Professor Hupy demonstrated with his own Y6 multicopter, shown in figure 16.

Figure 17 - Mission Planner software that has the checkpoints
and path for the Y6 to travel. On the left is a heads-up-display
that shows the altitude, pitch, and various other information
about the status of the Y6.

Before any mission, the mission must be planned out using a program that can communicate with the aircraft's on-board navigation system. There are many free, open-source programs that can accomplish this; in this case, we used Mission Planner (Figure 17). In this software, the path that the UAS travels is programmed using checkpoints over a map of the study area, as well as the elevation that the craft travels at. Both the checkpoints and elevation are extremely important so that the copter does not fly straight into a tree or a wall. When setting up the path, you have to know what the tallest obstacle in the are is so that you can set the elevation well above that. In this case, it was not too difficult since we were in an open field.

Figure 18 - Our class gathered around Joe Hupy
and his mobile command center (otherwise known
as a recycling bin).

Once everything was ready, the Y6 was launched. With Professor Hupy at the helm using his mobile command center (Figure 18), we watched the Y6 fly its course. Most rotocopters, including this one, have a battery life of only about 15-20 minutes. This limits how large of an area the UAS can cover. If it is a windy day, as it was the day we were using it, the battery life is even less since the craft has to use more energy to stay stable. The Y6 was not in the air for very long, but it was long enough to complete the path that Professor Hupy had programmed. After a few minutes of flight, the Y6 returned to the starting location and landed automatically.








Results

Figure 19 - One of the images taken by the Y6. Most of the
other images were unusable.

The results from the Y6 were not good, unfortunately. The images were very dark, and did not overlap well enough to be mosaiced. I tried mosaicing just like with the balloon data, and the result was just a dark blob. Figure 19 is one of the images taken, which is one of the best ones of the bunch.

Conclusion

Both the balloon and the Y6 had their pros and cons. As far as results, the balloon did much better than the multicopter. We got plenty of good, usable images from the balloon, but the Y6 didn't do very well. However, the image quality problem should be a small, easily fixed problem. The craft itself preformed exactly as planned. The Y6 flew an area of similar coverage as the balloon, but did it in under 15 minutes. It took us about 45 minutes to walk the fields, not counting the time it took to inflate the balloon and attach the cameras. The Y6 is more expensive and more complicated to operate, but is faster and under more control. The balloon is pretty cheap, and can still collect solid images, but is heavily influenced by weather and is far more time consuming.

The fact that we used a balloon at all was dependent on the weather. There was very little wind that day, but if it had been windy we would have used a kite instead. The Y6, and rotocopters in general, have far more practical uses. Consider if figure 19 was a house in an area experiencing severe flash flooding. Using a rotocopter to zip over areas and see where people are still trapped on rooftops would be extremely helpful, and it wouldn't put additional life in danger. That is not something you can do with a balloon. These assignments really showed the advantages and disadvantages of both types of UAS.

Used a balloon because it was not at all windy. If it was windy, we would have used a kite.
Simple technology- once it's up, just have to walk around, not much else you can do.