USE OF IMPROVISED REMOTELY SENSED DATA FROM UAV FOR GIS AND MAPPING, A CASE STUDY OF GOMA CITY, DR CONGO Cung Chin Thang United Nations Global Support Center, Brindisi, Italy, Email: thang@un.org KEY WORDS: Unmanned Aerial Vehicle, aerial photo, Nikon 1, city mapping ABSTRACT: In late 2013, the United Nations, under the authorization of the UN Security Council, launched the first Unarmed Aerial System (UAS) in Goma to support peacekeeping activities in the Eastern Democratic Republic of Congo (DRC). The system consists of an Unmanned Aerial Vehicle (UAV), sensors (cameras), a Ground Control Station (GCS) and control link (data link). UAVs, while generally associated with the military, have inspired a rise in civilian use, and the Geospatial Information System (GIS) Unit of the United Nations Organization Stabilization Mission in the Democratic Republic of the Congo (MONUSCO) saw the mission s new UAS capability as an opportunity. Compared to satellites, the technology offers rapid deployment and faster delivery of information at a much lower cost. It is also capable of surveying large areas very quickly and can be easily directed toward different regions on demand. Aerial photos were taken from a consumer grade digital camera mounted underneath the UAV and processed. Ten cm resolution mosaicked seamless aerial photo of Goma city was produced from 329 images. 1. INTRODUCTION United Nations Organization Stabilization Mission in the Democratic Republic of the Congo (MONUSCO) tirelessly work to restore law and order to support the Government of the DRC in its stabilization and peace consolidation efforts. Stability in the Western part of the country has been achieved but the East is still marred with several armed groups fighting for control over vast mineral resources. MONUSCO redeployed its main task force to the East in early 2014 to tackle instability in the region in a joint effort with the Force Intervention Brigade (FIB), created in 2013. They engaged various armed groups who were terrorizing the local population in North and South Kivu and Ituri districts with tremendous success and ultimately they were driven deeper and farther into their hideouts which became more difficult and dangerous to monitor their activities by conventional aerial reconnaissance. Eventually, MONUSCO contracted external operator SELEX ES from Italy for aerial surveillance using Unmanned Aerial Vehicle of FALCO model since late 2013. 2. MATERIALS AND METHODS 2.1 Objectives The primary use of the FALCO UAV in MONUSCO is for aerial reconnaissance with multiple sensors devoted to surveillance activities where the UAV maintains focus on selected areas and collects information about movements of people and vehicles. The information is available to the operator in real time and a live video stream is generated and disseminated as required. Although the existing onboard sensors (Electro Optical /Infrared, Synthetic Aperture Radar) were not considered optimal for mapping purpose, GIS office considered the UAV as an opportunity to use as carrier for remote sensors and hence conducted a test flight. The objectives for UAV test flight were to; determine if a UAV designed for a different operational purpose could be equipped with a digital camera for taking vertical aerial photos; capture aerial photos and process them using standard remote sensing/gis software; explore the potential use of deploying a UAV with integrated sensors for rapid response. 2.2 Methodology FALCO UAVs are capable of multiple payload installation and are equipped with full motion video sensors (EO/IR) within the main payload bay with Synthetic Aperture Radar in the nose. Instead of using the comparatively low resolution video sensors for mapping purposes, higher resolution still images acquired from a vertical-looking camera is considered much more useful and hence a consumer grade digital camera (Nikon 1-V2) with a 50mm lens was mounted with the lens facing downward within main pay load bay next to existing stabilized full motion video sensor.
Main Payload Bay Nose Payload Bay Figure 1. SELEX FALCO with multiple payloads capability (www.selex-es.com/-/falco) A FALCO UAV with Nikon 1 camera attached flew over Goma city at 1500m (4900 ft) altitude and above at 57 NM/hr on 7 March 2014. A total of 779 pictures were taken during the two hours flight time. Only minimal mechanic integration was performed on the aircraft and hence the consumer camera only runs on its own batteries (about 1.5 h from take-off). A total of 329 photos covering the main city area were, processed. Figure2. Goma City Flight Plan
2.3 Aerial Photo Properties As mentioned earlier, aerial photos were programmed for capture from consumer grade digital camera with a 1 inch sensor size and shooting was triggered by an external timer hence image acquision commenced before getting airborne, during pre-flight operations. Goma flight mission was able to achieve 10 cm resolution aerial photso using a lens featuring 18 degrees Field of View (FOV) while the UAV was flying at 1500 m above ground, with a shutter speed of 1/1000s to prevent image blur. Although the camera GPS recorded center point XY values of each image as EXIF embedded to JPEG, it was not very useful for automated mosaic in standard remote sensing/gis software. Furthermore, narrow FOV and small sensor size intended for covering one city (23.5 sq km area) requires hundreds of images to be mosaicked together. 2.4 Image Processing Remote sensing software automatically aligned and mosaicked images by reading systematically recorded locational information, either embedded in the images or stored in the header file. However, captured JPEG images did not contain such required information except XY coordinate for center point, therefore photo stitching software PTGui - was applied instead. The software generated a panoramic view by matching object edges of neighboring images rather than locational information. Since Goma APs were taken with 30-50% front and side laps at constant interval, it made the mosaic process smoother in PTGui software with marginal mis-alignment. The mosaicked AP was image-to-image georeferenced in ArcGIS against a recently acquired 0.5 m resolution WorldView-2 image with simple polynomial order processing. The result showed Root Mean Square Error (RMSE) of more than 5 pixels and therefore further correction was done by applying ground control points from Trimble GeoXH GPS, with less than 1 m accuracy. With GCPs, RMSE improved to 0.7 and a georeferenced Goma City AP mosaic was generated. We also tested creation of orthophotos using 30m ASTER Digital Elevation Model and the RMSE value further deviated away, this appeared to originate from big differences in resolution between AP and DEM (0.1m vs 30m). Figure 3. Image mosaicking process was performed in PTGui 3. RESULTS AND DISCUSSION The equipped test sensor was not meant for mapping purpose and required complex image processing compared to the capabilities of a professional mapping sensor, which would have stored valuable geo-information such as location, orientation, height, etc. Potential uses of the end product with its stunning 0.1m resolution are potentially limitless, not only for large-scale mapping (up to 1:2,500) but also in providing many details on the ground. It can also be beneficial for surveying and engineering works. MONUSCO Engineering Section was pleased with the AP mosaic from which detailed work can be planned, thereby reducing a lot of ground surveying.
Moreover, updating the Goma city map becomes much easier with stunning resolution enabling enriching existing features and adding more layers for value added products. One potential use of AP from UAV is better road classification such as detection of good condition roads which currently is conducted on the ground in the mission. Figure 4. Ground resolution of 10 cm (0.1m) aerial photo Figure 5. Comparison of Aerial Photo (0.1m) vs Pleiades Satellite Image (0.5m)
4. CONCLUSIONS GIS offices in DPKO field missions acquire various satellite images from service providers often involving a lengthy acquisition process which highly impairs clients needs in seeking timely information. Depolying a UAV with onboard mapping sensors onboard definitely has the potential for compensating geospatial information requested in the field mission with better spatial and temporal resolution depending on the UAV range and sensors specification. Although there might be several limitations for using AP acquired from a consumer grade camera, a lot improvement can be made by deploying commercial grade mapping sensors, which cover a much larger area and require less acquisition and processing time. Some sensors have ability to acquire both vertical and oblique photos (color and infrared) simultaneously which can be useful for military intelligence, allowing detection of partially hidden objects (e.g. targets hidden by trees) in addition to enhancing image interpretability. For example, the same point of interest can be observed from up to 5 different vantage points of view from Leica Geosystems RCD30 sensor. The additional advantage of using the UAV as a GIS tool is that it can traverse dangerous and inaccessible areas where fieldwork is difficult, thus reducing the time and expense involved in logistical planning of field missions. Since the aircraft flies at low altitude below the cloud cover, it has already proven ideal for challenging regions like the Eastern DRC. This test flight proved its usefulness, especially for mapping purposes and maps can be produced or updated with the most recent information on the ground which ordinarily was done rarely, due to the high costs involved. ACKNOWLEDGEMENTS We are grateful to the MONUSCO UAS Cell and SELEX ES Team for supporting the GIS Unit with aerial photos and other technical assistance during the flight mission and post processing of images. We also thank the Chief GIS and Chief CITS, MONUSCO for allowing us to conduct this research study for high resolution aerial mapping of Goma City. REFERENCES: FALCO UAS System, http://www.selex-es.com/-/falco ICT Insider Issue 20, United Nations HQ New York, ICT Division Quarterly Issued Internal News Letter PTGui, https://www.ptgui.com/