Activity 1: Image Gathering Fundamentals: Using a Kite (Weather Balloon) to Gather Aerial Imagery

Introduction

This was the first week of this Unmanned Aerial Systems class where we did a field activity. This week we focused on collecting aerial imagery in an old school fashion. This served the purpose of demonstrating to the students that a more complicated and automated Unmanned Aerial Vehicle or UAV isn't the only way that useful aerial data can be collected. In the next few weeks they will be working with and learning how the same techniques used in this weeks low tech activity also apply to the much more advanced and technical platforms available today for data collection.

Study Area

For this weeks activity we went to the Eau Claire Sports Center complex (Figure 1) which is about 5 minutes from campus. The complex has a large area of multiple soccer fields which was good for this activity, giving us plenty of space to collect data in. It was a cloudy day with temperatures in the lower 70s and very little to no wind. The original plan was to fly a very large kite (Figure 2) with the cameras attached but because of the lack of wind a weather balloon (Figure 3) filled with helium was used instead.

Figure 1 is a Google Earth image showing the sports center complex (lower left) and the soccer fields we used during this activity. The yellow box is the area that we walked back and forth over with the balloon to collect imagery. The huge amount of open space was ideal for this exercise. 

Figure 2  The original plan was to use a kite very similar to the one pictured here. It is about 6
feet across which helps with lift and makes it more stable.

Figure 3 This is a weather balloon similar to the one used in this activity.

Methods

The methods for this activity are fairly straight forward and not overly complex. Dr. Hupy assigned some pre class readings on how to set up the balloon with cameras using a device called a Picavet Rig (Figure 4). The purpose of the picavet rig is to keep the cameras facing down at all times to collect vertical imagery. The strings are tied in such a way that they are allowed to slide back and forth as the balloon or kite tips to keep the cameras 90º to the ground at all time (Figure 4). Dr. Hupy had a picavet rig set up from past activities that we used for this activity however he showed the students how it was strung up to help them understand the concept. Once the rig was set up the next step was to inflate the balloon. The balloon was filled with helium which is lighter than air so that the balloon would float. We had to put in enough helium so that the balloon wouldn't only float but be able to carry the two SX260 Canon cameras (Figure 5) mounted on the picavet rig. There are two cameras because one is a regular true color or RGB camera and the other is an infrared camera. This will allow use to look at the area in normal color but also in infrared from which we can do vegetation health analysis. Once the balloon is inflated a string is attached so that we can control how high the balloon goes and where it goes. The picavet rig is attached to the string about 15 feet below the balloon. For our purposes the balloon was flown at approximately 50 meters or 150 feet. This is a good height which allows use to get the desired amount of overlap between images. Image overlap is essential for creating accurate and usable data layers such as a mosaic of the whole area where data is collected. 75 % overlap is a good amount, but during this activity we were getting around 80 to 85% overlap. Again that overlap comes from flying at an adequate altitude and well as the correct spacing between passes. Data was collected by walking straight lines back and forth across the soccer complex and keep the balloon at a constant altitude and keep the same spacing between passes which in this case was 10 meters or 30 feet. Figure 6 is video of the balloon while it was collecting imagery.

Figure 4 This is a diagram of how to string up a picavet rig. It is very simple to set up and is made of simple string, a metal cross piece and some special clips used to attach it to the string on the balloon.



Figure 5 This is an SX260 Canon camera like the one used in this activity. It is a fairly inexpensive camera that takes good quality images and is easy to use. This camera is bolted to the metal plate on the picavet rig 90º to the ground or the lens pointing to the ground.

     Figure 6 This is a video Dr. Hupy took with a DGI Phantom which is a small drone with a very good camera on it for collecting video and images. You can see the weather balloon and the cameras attached to the picavet rig.

Discussion

Later in the semester we will be processing the imagery we collected during this activity and it will be interesting to see how the data turns out. With a multi rotor or fixed wing platform the amount of movement side to side the cameras do is pretty minimal compared to a platform like a balloon where the picavet rig and cameras can sway side to side quite a bit, The picavet rig is supposed to minimize the movement but we will see how well it did when we process the data. Another consideration when using this platform compared to a more sophisticated one is the time between images. With a mulitrotor where you are moving at 15 to 20 MPH you will want pictures taken more often usually around every .7 seconds, with a slow moving balloon that can adjusted to every 3 to 5 seconds or even longer depending on how fast you are walking. The speed and timing of image capture is crucial because just like we want 80% overlap side to side on the images we also need the images to overlap front to back a decent amount. Again it will be interesting to see how good our timing and overlap was when the data is processed in a couple of weeks.

Processed Data

I ran the imagery data we collected during this exercise and it turned out pretty good. Figure 7  is a 3D mosaic of all the RGB photos taken during the exercise. There were 570 photos that I brought into a image processing program called Pix4D, which takes all those individual JPG files and combines them into one large image. It does this by using the geotags that the camera places on each image. A geotag is a lat/long position where the image was taken. Figure 8 is the same mosaic but you can see in the red circles there are some areas of the image that the program didn't mosaic properly. There are a large number of reasons why this happens, both data collection methods issues as well as software errors. In this instance you can see some issues with elevation. Just as the camera takes a geotag it also records an elevation value for each image. If these values get messed up you will get areas in the image like the red circles where this area is obviously flat but the mosaic has drastic elevation changes in it. These errors can be erased so they aren't too big of a deal. For being attached to a balloon this data turned out extremely well. Figure 9 is a 2D version of the mosaic brought into ArcGIS where it can be manipulated and have various GIS tools and processes run on it, such as a simple elevation map Figure 10. Along with these RGB images there were also 570 Near Infrared images taken that I am currently working on mosaicking. From what I can tell so far the mosaic isn't going to turn out as well for the IR but will have to wait and see till it is done processing.

Figure 7 This is the 3D mosaic created in Pix4D of all the RGB JPG files taken with the weather balloon of the Eau Claire Sports Center soccer complex.

Figure 8 This is the same 3D mosaic but you can see the errors discussed above in the post.

Figure 9 This is the 2D version of the mosaic brought into ArcGIS.



Figure 10 This is a very simple elevation map created form the DSM file which is the elevation file created when the images were run in Pix4D. The green areas are higher elevation and the red areas are lower elevation. The green area circled in black is one of the errors seen in the 3D mosaics above.



Conclusion

This was an interesting exercise for me. I am quite familiar with collecting aerial imagery using UAVs, as I am the technician for the Geography departments UAV lab and program, however I have never collected imagery in this low tech way. All of the concepts are the same it just comes down to making slight adjustments for the difference in platform technology, speed and stability. It is good for the students to see this method before they see the more high tech platforms later in the semester so that they realize that collecting aerial imagery can be done cheaply and effectively without using a platform that costs thousands of dollars. It also gives them a good idea of how drastically the technology has advanced in this area over the last couple years.