This week we managed to collect another data set of diurnal heat patterns. However, we ran into several problems that bring to light a couple of the difficulties of working with UAS and thermal data.
The first problem we encountered was that the planned location of data collection, the Tippecanoe County Amphitheater, would be displeased at the prospect of us bringing the m600 and collecting data throughout the day without first contacting them. We decided that to avoid possible confrontation, that we would relocate to the Davis Ferry location right up the river.
We began our first flight from the red circled area and captured data in the red boxed area.
 |
| Davis Ferry |
The amphitheater location is still on the cards, we will just have to get their permission beforehand while also making sure that the date works for all group members involved, has winds below 18 mph all day, the sky is clear enough that there are discernible differences between data sets, and that nobody else is using the m600 on that day.
 |
| Sunrise data set |
 |
| Morning data set |
The very first flight, at sunrise, can be seen on the left. The second flight, at 11:30 can be seen on the right.
The combination of the innate low resolution of the thermal camera and the large amount of very uniform water proved to be too much for Pix4D to calibrate most of the cameras.
At that point we realized that the current data collection methods would not work for this area. We had to switch over to a method where we flew one flight at 90%, and then flew a perpendicular flight also at 90% resulting in about twice the number of images collected. This also had the effect of doubling the processing time. The resulting data set came out looking much more dense but resulted in a usable orthomosaic.
 |
| Afternoon (4:30 PM) data set |
 |
| Sunset (7:00 PM) data set |
Another problem that we encountered was that we would lose georeferencing on the thermal data set when the m600 was in an area in the top right of the collection route, the furthest point from where the pilot in command was standing. What this meant was that we were not able to use the images collected in the top right portion when processing in Pix4D. The RGB data sets, however, did not experience this problem.
 |
| RGB has full georeferencing |
As a result of these problems, we were only able to gather usable thermal data from the afternoon and the sunset flights.
The afternoon thermal orthomosaic appeared to be cut into pieces and the resampling type had to be changed to cubic to be fixed.
 |
| Afternoon shifted |
 |
| Resampled Afternoon |
The sunset orthomosaic came out jagged and malformed. There are erroneous cold spots on the edges that skew the values for the rest of the ortho.
 |
| Sunset |
We set out to explore the strengths of using thermal imaging with UAS but we instead stumbled upon one of its greatest weaknesses. Flying over water or dense vegetation introduces new problems with processing an orthomosaic that can be mitigated but not completely resolved.
Here's the morning orthomosaic.
 |
| Not at morning person, clearly |
Here's a swipe map of the first Purdue Wildlife Area data collection test. We were considering including something like this in our poster (via QR code) but have yet to figure out how to properly implement it.
