Svarmi grew on all fronts in 2018. Not only have we completed more mapping projects than any previous year, but we have also had huge developments which will hopefully put us on the map as innovators in the field of remote sensing with drones. Only two months in, 2019 is already looking like a bright year at Svarmi. Here are some things we have achieved this year in more detail, plus some things we look forward to in 2019.
Mapping Projects in 2018
Thermal, LiDAR, and more
In 2018, we mapped over 155 sq. km (0.15% of Iceland!) in nearly every corner of Iceland, and used every drone and every sensor that we own to do so. Therefore it goes without saying this year did not disappoint in terms of project diversity. Here we discuss a few highlights of projects which will continue in 2019, as well as list some new devices and applications of drone remote sensing that have been developed this year.
<< A map of the drone mapping projects at Svarmi in 2018. Dots indicate mapping projects, the color of the dot represents the type of sensor used to map: purple = color (RGB) camera, yellow = LiDAR sensor, blue = multispectral & thermal cameras, red = thermal camera. Imagery: Landsat8 natural color mosaic (Landmælingar Íslands 2014).
One of our favorite projects this year was mapping the area affected by the August 2018 Skaftárhlaup in collaboration with Icelandic Meterological Office and Vegagerðin this fall. In this project, we mapped nearly 45 sq. km in high-resolution of the area affected by the flooding. This project was challenging due to its size, but rewarding. As a result of this project, we got some amazing imagery of the flooded area and an opportunity to work with and learn from experts in geohazard mitigation in Iceland. This data will be used to create a hydrological model by experts at Veðurstofa, and we greatly look forward to similar projects in the future.
Svínafellsheiði is an excellent example of the uses of drones in geo-hazard monitoring in remote locations. Daniel at Svarmi has been working with Þorsteinn Sæmundsson (HÍ) to monitor this area with drones since 2016. However, the area attracted media attention in 2018 when researchers discovered that a crack in the mountain side was many magnitudes larger than previously thought. Additionally, the surprising seismic activity in Öræfajökull in 2018 raised questions about the stability of the area, putting the Svínafellsheiði project at even higher significance. The area was closed to guided tours in the fall, and Svarmi continues to work with HÍ and Veðurstofa Íslands to monitor the movements of the crack.
A 3D model of Svínafellsheiði from flights this summer of the area. >>
Finally, we continued work on the ongoing project on Skeiðarársandur flood plain directed by Kristín Svavarsdóttir at Landgræðsla Ríkisins and Þóra Ellen Þórhallsdóttir at Haskóli Íslands. In this project, we use a multispectral camera, thermal camera and ground-based measurements to evaluate the downy birch population in the area. We returned to Skeiðarársandur 3 times during the summer to repeat high-resolution surveys of the area, providing time-continuous data. This project highlights the strengths of drone-based remote sensing for research: it provides a vast amount of information at a lower cost than ground-based surveys. We look forward to continuing this very interesting project next year. Furthermore, Victor at Svarmi will continue to use machine learning to classify plant species on the floodplain and understand the distribution of birch from year to year.
What is LiDAR?
Svarmi is now a proud co-owner of the first drone-based LiDAR (Light Detection and Ranging) sensor in Iceland (along with Verkís). LiDAR works by measuring the time it takes for lasers of light that it sends out to return to the sensor, giving a very accurate elevation model of every point on the ground. Our LiDAR sensor is a Riegl Mini-VUX1 which is mounted on a Matrice 600 (DJI) drone. Needless to say, we are very excited to be the first Iceland-based LiDAR operator.
<< A close-up view of the drone-based LiDAR sensor – the only one of its kind in Iceland.
One main advantage of LiDAR over photogrammetry is that the sensor can detect surfaces in all light conditions (even at night). A second big advantage of LiDAR is that the lasers sometimes hit the ground below vegetation due to multiple signal returns. As a result, LiDAR can provide better terrain models (DTMs), or models of the ground surface below the vegetation cover, than can be created with photogrammetry. Finally, LiDAR points are sent 360° while the drone is flying over an area of interest, meaning that, unlike photogrammetry, it can be used to map vertical structures such as power lines and walls quickly and accurately. Because the lasers sent out by the LiDAR unit reflect off of surfaces at various heights, LiDAR is particularly useful for forestry or agricultural studies. For example, LiDAR can be used to find the height of the vegetation above ground, the amount of leaves on the vegetation, and the various levels of vegetation (e.g. trees vs. bushes vs. grass).
Thermal Mapping with Drones
Using a fixed-wing drone equipped with a LWIR (long-wave infrared) thermal camera, we have successfully completed several mapping projects this year. This type of mapping method has many applications, but this year we mainly used it to map geothermal areas which could be developed in the future and to inspect pipes carrying hot water for leaks and damage. Drones which are equipped with thermal cameras are becoming easier and cheaper to operate, but often thermal cameras that come standard with them are not designed for mapping purposes and as a result produce large temperature drifts or inaccuracies. At Svarmi, we use a mapping-grade LWIR camera which is designed specifically for mapping at high accuracy. Additionally, using a high-quality thermal camera allows us to create absolute temperature maps of the surface, rather than relative temperature.
<< Steytingur (fixed-wing drone) with a thermal camera at Reykjanesvirkjun this summer.
Applications of drone-based thermal mapping in Iceland
This year we made large developments in our efficiency and accuracy at mapping large areas a thermal camera. Perhaps the most obvious application of this method is to map and monitor large geothermal fields, or for identification of geothermal areas which have not yet been explored. This year we had the privilege of working with two of the biggest names in geothermal energy in Iceland: HS Orka and Reykjavík Geothermal. Our collaboration with these two companies helped us to improve this drone-based remote sensing method for geothermal exploration purposes.
We custom-built a fixed-wing drone with a thermal camera in order to map large areas (e.g. geothermal fields). In one example, we created a high-resolution thermal map of Reykjavesvirkjun and surrounding area for HS Orka. Fixed-wing drones can map larger areas in a shorter amount of time than commercially available multi-copters fitted with thermal cameras. For example, in one 50-minute flight, we acquired over 1500 thermal images. From these images, we created a large thermal map (several sq. km) at 40 cm /pixel resolution; this is between 100 – 1000 times higher resolution than is available from thermal satellite imagery today. Additionally, thermal drone images are less affected by atmospheric interference (e.g. from clouds) than satellite images.
Large pipeline systems carry hot water and steam all over Iceland. Some of these are more than 50 years old and require constant maintenance. Both internal and external corrosion can cause leaks to the pipes, but since the pipes are cased in thick insulation it can be difficult to see internal leakage from the outside. Therefore, HS Orka set up a pilot project in collaboration with Svarmi to find out whether leakages and other failures can be detected through thermal anomalies in drone images.
A thermal image of a geothermal pipe. Brighter colors represent warmer temperatures, darker colors represent colder temperatures >>
An example of a leak identified at Reykjanesvirkjun using thermal drone images. The strong thermal anomaly can be seen in the thermal image (left) as a brightly colored spot near the joint of the pipe. The RGB image of the same location shows no damage or leak at the surface (right). Bright colors represent warmer temperatures, dark colors colder temperatures.
This project is the first-ever thermal drone inspection of geothermal infrastructure in Iceland. In order to detect thermal anomalies in the pipes, one of Svarmi’s multicopters fitted with a thermal camera flew at low altitude over the pipelines, gathering thousands of high-resolution thermal and RGB images. From these images, digital thermal maps of the pipelines were created. Thanks to Svarmi’s thermal drone survey, at least one major internal leakage has been found. This leak would not be possible to identify from simple visual inspection of the outside the pipe, but with the thermal camera a strong anomaly can be seen. Due to the accurate geolocation of the thermal map, it was convenient for the technicians on site to quickly find and repair the leakage, thus saving many hours of manual searching in the field. Furthermore, drone thermal inspection greatly increases the safety of the staff on-site, since the time that staff members spent in the hazardous areas near the geothermal pipes was reduced considerably.
Thermal Mapping in Ethiopia
This winter, we got to test our methods and equipment outside of Iceland in one of the most challenging (yet rewarding) projects of the year: mapping a geothermal area in Wolaita Sodo, Ethiopia. This mapping project was part of an ongoing geothermal research by Reykjavík Geothermal in the area. Although the flying conditions couldn’t be more different than in Iceland, we managed still to map a large area which is being currently researched by geologists at Reykjavík Geothermal (RG) as a potential site for a power plant. For this project, we used again a fixed-wing drone fitted with a LWIR (thermal) camera. We are currently reviewing the thermal data that came from that trip with the geothermal specialists at RG to further refine the method for use at lower latitudes so that the imagery can be useful to the geologists.
Svarmi’s fully-automated drone program
What is Myriad?
In the Myriad project, we are developing a solution to provide drone data as a service using a fully automatic drone. Therefore in 2018-2019 Svarmi has focused on setting up and solidifying a secure and efficient software infrastructure which will receive and process data from hundreds of Myriad drones that will be deployed to industrial areas in the future. The Myriad drones will communicate to the software infrastructure through the 3G/4G/5G network. Along with setting up the required back-end software infrastructure for the system in 2018, we have also significantly developed our web dataviewer which allows users of the Myriad system to easily inspect fully processed data from our drones and interact with the system.
A visual explanation of myriad >>
Svarmi’s Myriad project is funded by Rannis’s Technology Development Fund (TDF). In 2016-2017 Svarmi developed a prototype of a Myriad drone with the “Sproti” grant from TDF and for 2018-2020 Svarmi got the “Vöxtur” grant from TDF which allows Svarmi to further develop and solidify the solution developed in 2016-2017. The software being developed in this project will be used also on the drones operated by Svarmi staff (not fully automatically) to speed up and improve data processing.
<< Svarmi programmers Aðalsteinn (left) and Hallgrímur (right) work on the Myriad live data streaming solution.
The 2018 ESA Copernicus Masters is a global competition about problem-solving using earth observation data. We entered Myriad, Svarmi’s automated drone project, in the ESA EO Future Challenge Category of the competition. As a result, we are happy to announce that the 2018 ESA (European Space Agency) Copernicus Masters committee chose Myriad as a the winner for this category. Additionally, the winners of the Copernicus Masters receive access to big data from the ESA Copernicus Program. Above all, access to these data opens many doors for the future of Myriad at Svarmi in terms of environmental monitoring with drones and satellites.
As a winner of the Copernicus Masters Challenge, Svarmi will have credits to use DIAS (Data and Information Services) for Copernicus satellite data. This is useful, for example, in processing large numbers of satellite images. Having DIAS access will be a huge leap forward in the development of Myriad. For example, we can now improve our ongoing methods of environmental change detection can be up-scaled from drone images to satellite images. As a result, this will create more accurate and complete data than are currently possible using either drones or satellites for monitoring environmental changes.
Machine Learning & Remote Sensing
Lupine Classification in Iceland
Victor has become a machine learning expert at Svarmi over the last year or two, and has developed techniques to use Satellite and drone images to classify vegetation. Using TensorFlow and PyTorch, he has developed sophisticated machine learning algorithms which are useful here in Iceland. For example, he has been working with Landgræðsla Ríkisins to estimate populations and distribution of native species like Downy Birch (Birki), as well as invasive species such as Lupine (Alaska Lúpína). Using a CNN (convolutional neural network) which Victor designed and trained himself with ground-truthed data, he has been able to estimate species populations with great accuracy (up to 96% in some cases). Furthermore, the CNN can be expanded beyond just vegetation classification: it could also be used, for example, to identify hotspots in thermal images, estimate vegetation mass, and automatically detect environmental changes. We very much look forward to seeing more results from this exciting technique and hope to get more projects and see more applications of the technology in the future in Iceland.
Other Applications for CNNs in Iceland
Another example of the application of machine learning in remote sensing is an ongoing project at Svarmi to automatically classify seaweed beds in Breiðafjörður based on satellite images and up-scaled drone images. Victor has been working on this project in connection with Hafrannsóknarstofnun since 2016 and will continue to refine the method in the future. CNNs can be expanded beyond just vegetation classification: it could also be used, for example, to identify hotspots in thermal images, estimate vegetation mass, and automatically detect environmental changes. We very much look forward to seeing more results from this exciting technique and hope to get more projects and see more applications of the technology in the future in Iceland.
All in all, 2018 was a great year for Svarmi. Just to list a few accomplishments: progress in the Myriad automated drone project, a big award from the European Space Agency, a new drone-based LiDAR, developments in machine learning for remote sensing, more mapping projects than ever before, and two new team members. Thank you to everyone we have had the pleasure of working with in 2018, we could not have improved without you. We very much look forward to more projects in 2019.