POB asked Art Kleinrath, Department of Energy Office of Legacy Management (DOE-LM) program manager, and Kenneth E. Karp, Navarro Research and Engineering Inc. senior program manager, along with Jackie Petrello of Leidos and Seth McBeth of Aero-Graphics to describe the planning that went into a thermal imaging project at remote sites in Idaho and Montana.
The DOE-LM needed to assess environmental liabilities at sites where non-nuclear explosives tests were conducted in the 1950s and 1960s under the U.S. Atomic Energy Commission projects from the Plowshare and Vela Uniform Programs. The Plowshare Program was established to develop peaceful uses for nuclear explosives. The Vela Uniform Program was designed to provide information to improve the capability of detecting, monitoring and identifying underground nuclear detonations. As a result, Navarro Research and Engineering Inc., contractor to DOE-LM, selected five Plowshare and Vela Uniform sites for evaluation. Two of these sites were candidates for aerial surveys. These two sites were too large for effective and efficient ground-level visual inspections, and contained features of interest that could be identified from aerial photography such as craters, cleared areas from project operations and large pieces of debris.
The scope of the project included collecting aerial photography and thermal data at two of the sites. The survey data was compared to operational histories and site visit observations to identify whether there are remnants of the former activities at the sites that could represent potential environmental liabilities such as structures, craters, storage tanks, wells, debris and drilling mud pits. Navarro used this information to provide recommendations to DOE-LM for future actions at the sites.
Despite previous experience with thermal acquisition from a manned aircraft, Aero-Graphics had never done thermal acquisition from an unmanned aerial system (UAS) platform. One of the first decisions was what kind of thermal camera to use. During the planning phase of the project, the team spoke to many different thermal camera manufacturers and reps to get as much information about the capabilities and limits of sensors sized for UAS use.
The first consideration was whether to use a vertical-take-off-and-landing (VTOL) platform or a fixed-wing drone for acquisition. The benefits of a VTOL drone came from having a gymbal-mounted camera as opposed to one that was hard mounted. This offered more steady and higher quality imagery. The drawbacks came in the size of the project — close to a square mile of acquisition. This raised concerns the relative temperature of the ground would change enough through the course of acquisition that, despite the higher image quality, the VTOL drone would produce a lower overall quality of data unless flight times were restricted. This would add up to two days to both sites. With the fixed wing that was ultimately selected, the thermal camera was hard mounted to the frame of the body, so the camera was susceptible to the same bumps and vibrations that the air frame of the drone encountered. However, it was able to potentially cover RGB and thermal acquisition in the same day. Further research into processing of the thermal camera/fixed-wing combination showed it would take thermal imagery, in somewhat of a time lapse fashion, up to six frames a second, and then reject up to 90 percent of the photos based on image quality. This would help reduce the number of blurred images from the hard mounted camera. With the equipment decision made, the team fell back on most of the same resources for any other manned or drone acquisition.
After that, it was time spent studying the layout of the terrain, the overall size and shape of the project, and where the best possible spots were to place control targets. Considerations included both the survey aspect of the job (to ensure proper coverage of control points) as well as what made the most sense economically. Access was limited by the rough terrain and most of the targets were going to have to be walked to. Some of the other resources used included bringing a certified land surveyor as the spotter for the project. By using a crew that was trained in land surveying as well as operation of the drones, it was possible to set targets efficiently and get a general idea of the landscape the day prior to acquisition, and then start right into acquisition the next morning.
Though many aspects of the project were familiar to Aero-Graphics, there were some twists, especially when planning a drone mission so soon after part 107 was released. The logistics of coordinating special activities permits to fly a UAV over Army Corps of Engineers land, coordinating with local wildlife rangers to fly within their wildlife bird refuge, coordinating with local range control to fly within military operations areas, and making local FSDO offices aware of project operations were part of that. In terms of conventional survey and the role it played in this project, once those issues were dealt with, targets were laid on the first day and static base stations set through the project. Roving from one target to the next, the necessary information for control points was collected. Once all that information was collected, the focus shifted to the drones and acquisition until the data was taken back to the office for processing.
Different from Manned Operations
With manned aircraft acquisition, planning is less complex because, in many ways, the aerial and terrestrial aspects are more consistent. With UAS operations, local weather conditions and wind speeds/temperatures become much more crucial. You use as many local weather sources as possible to start building a general sense of the area to be able to predict wind conditions at certain times. Because the acquisition was mostly focused on thermal imagery, there was a lot of consideration put into how the atmosphere and the ground were going to be heating up through the course of the day. The thermal imagery was going to be used to pinpoint spots on the earth that were emitting higher temperatures than the surrounding ground. So initially the idea would be to acquire the imagery as close to sunrise as possible to map the ground during its coldest part of the day and provide the highest contrast while still being able to satisfy the daytime regulation rule of part 107. The material the ground control targets used relied on the same radiation from the sun to heat up and become visible on the thermal imagery. One solution was laying bags of ice over the target to give off a colder, darker appearance than the ground. Another alternative was a combination of warmers and battery-powered heaters to help the targets appear hotter than the general area. The logistics of hiking to each of the targets to either heat them up or cool them down became enough to consider other options. In the end, a series of experiments were done before the project using different combinations of materials that would contrast against themselves. Many early mornings were spent measuring the temperatures of the target combinations as the sun heated them up and flying test imagery to study the visual differences between the materials. The solution was to rely on a 25-degree sun angle to start acquisition. This would be a sufficient amount of time to heat up the targets, while still being early enough for the ground to remain cool.
The diverse terrain between the two sites affected how the surface of the ground responded thermally. In the case of the Montana site, the terrain was generally muddy and wet, comprised of many puddles of water due to the high content of clay in the soil. With little brush and grass growing in the area, it proved ideal when mapping the temperature of the ground. The Idaho site was mostly rolling hills covered in tall grasses. It proved extremely difficult to calibrate the thermal data because of how the grass acted as an insulator. Because the mapping included using RGB imagery and photogrammetric techniques, it was very difficult to generate a true surface of the earth, as most of the points sat on the tops of the grass well above the actual ground.
Most of the project went off close to how it had been planned. The data results support the decision to use a fixed-wing drone, though a larger more capable drone could have made the project much easier. Given the option, using a manned aircraft for LiDAR imagery or a drone-mounted LiDAR unit on the Idaho site could provide the best surface data. Both locations were much more remote than anticipated. Looking back after having driven the roads, mobilizing to the site in two vehicles instead of one would provide more options in case the first vehicle got stuck.
The final deliverable consisted of:
- RGB and thermal orthomosaics of both sites
- DSM and point cloud data for both sites
- Thermal video of acquisition
- All associated pictures of the project on site
Using the systems described achieved a 1.3-inch GSD of the RGB imagery, and a 1.6 GSD of the thermal, due to the lower quality of the camera.