River Mapping Using Low Cost Unmanned Aerial Vehicle with Remote Sensing Platform

Yiding Han

1. Introduction

1.1 Outline

The mapping of river paths is of great significance in GIS. However, the cost of the current remote sensing systems, such as satellites and manned aircrafts, for river mapping is extremely high. Updating geographic information with current systems normally takes several years. The process is tedious, and even has a high-risk of endangering human lives. Since part of river path changes seasonally due to flood or drought, the current system cannot provide sufficient information for corresponding geological analysis.

Therefore, the principle goal of this research is to develop a low lost fixed-wing Unmanned Aerial Vehicle (UAV) with remote sensing platform to replace the current system in certain circumstances. A UAV remote sensing platform has many advantages over the current systems, not only by its lower cost and better employability, but also higher resolution images. The UAV navigates at a much lower altitude of several hundreds of meters, which enables it to carry much cheaper and less complex remote sensors, such as digital CCD cameras, to take images with fine detail.

A map of the river is produced by stitching the images together. The georeference data, such as the aircraft flight data and GPS information, is logged for each picture. Thus, every image can be projected onto the actually earth map. With certain overlapping area (>30%) between the adjacent pictures, the projected pictures can be stitched together to form a map of the river

.

Fig. 1a original picture

Fig. 1b orthorectified picture

1.2 System Introduction

1.2.1 Unmanned Aerial Vehicle

UAVs are unpiloted aircrafts which can either be controlled remotely or fly autonomously. Although UAVs are well known for their military roles including reconnaissance or attack, they are also used in a growing number of civil applications, for instance in forest firefighting where it is too dangerous for a human observer, sending UAV into the fire zone can provide the clear and accurate information of the situation, without risking any human life.

1.2.2 Imaging system

The UAV platforms that is presented in this paper is a fixed-wing autonomous UAV with multi-spectral imaging system called the GhostEye. GhostEye is a payload of the UAV that provides imaging ability to the system. The cameras that are used in the system are normal CCD cameras which have the ability of remote capturing, meaning that the camera's functionalities are possible to be accessed from a computer. The imager can be both RGB and NIR cameras; also the number of the cameras loaded on the platform is not limited. In addition, while taking images of the ground, the flight data, including roll, pitch and yaw angles of the UAV is recorded, as well as the GPS data of the UAV. The data is timestamped with each picture and stored in a SD card. Subsequently, the airborne images can be well georeferenced and orthorectified. Therefore, the orthorectified images are to be stitched together and form a map of the river path.

Fig2. The UAVs and their onboard camera

1.2.3 Flight path planning

Pre-programmed flight plan is needed in order to fly the UAVs autonomously. The flight path planning is essentially designing way points which are followed by the UAV during its flight. The way points are defined as coordinates with given longitude, latitude and the altitude, saved in the format of XML file. Hence, the way points are readable by both GCS (Ground Computer Station) software and the UAV on-board piloting processor.

Fig3. Flight path planning in GCS

2. GIS Based Flight Path Planning

2.1 Objective

Way point planning on GCS is time consuming and inaccurate. Therefore, in order to create a better way point generation solution, a GIS based flight path planning method is proposed. For river mapping mission, the available hydrological information can be used to create waypoints. The following procedure is suggested:

1. Download the river flowline data from NHDPlus, DEM from USGS.

2. Load the data into ArcGIS and project the flowline and DEM on the same map.

3. Choose the river and create a new layout. Select the part of river for the mission.

4. Use statistic tool to verify the flight length.

5. Output coordinates of the turning points of the flowline, as well as the corresponding elevation value.

6. Use a C program to read the coordinates and save as XML formation.

2.2 Data source

Our flight test site is in Cache junction, UT, where Bear River flows nearby. To test the GIS based flight path planning, the flowline shape file of Bear River is downloaded from http://www.horizon-systems.com/nhdplus/HSC-wth16.php.

Also, DEM of cache junction area is downloaded from USGS http://seamless.usgs.gov.

2.3 Create the flight path map

In this mission, the distance of in the projection requires better accuracy. USA Contiguous Equidistance Projection is used to create the map. Use the selection tools that are provided by ArcGIS, part of Bear River is selected and created as a layout for the river mission. The layout is plotted along with the DEM of this area.

Fig4. Flight path map created by ArcMap

To design the optimized flight path for the UAV, it is important to know the total travel distance. This is easy to find out using the statistic tool by viewing the attributes table of the river segment. It is shown that this part of river is about 13945 meters long. On a clear and calm day, our UAV’s navigation distance is about 36000 meters, which means travel distance is safe for the river mapping mission.

2.4 Output the waypoints

For flight path planning purpose, the waypoints that need to be generated should be above the flowline. Since flowline is defined as polylines in GIS, it is intuitive that the nodes of the polylines should be the waypoints for UAVs.

By converting the polyline to raster and then to points, we can generate all the points along the river flowline. Pick out the points that are on the node of the polyline and those which appear to be important for the flight mission and create a new layout. Add x y coordinates to the attribute table of these points, also the DEM value of these points. Subsequently a table of 3D coordinates of these points is created. Export the table and save it for further process.

A simple command line C program is made to generate the waypoints’ xml file with the given table of 3D coordinates and the height that the UAV flies above ground level. The program is simple and only based on ANSI C, so that it can be used on different operating systems, Linux, Windows, etc. The coding is simple, so it will not be discussed in this paper.

2.5 Edit the waypoints in GCS

The waypoints generated from ArcGIS are essentially the coordinates of the river path. However, some of the corners in the river path might be too sharp for the UAV. In order to avoid these aggressive sharp turns, we need to edit some of the waypoints manually in the GCS program.

Fig 6. The generated waypoints in GCS

To avoid the sharp turns, the UAV is sent into a larger circle that turns in the opposite direction of the sharp turn. An illustration is shown in the following plot.

We can see that waypoint R27 is not on the river path. This is because GCS flight simulation program has reported that the turn at this point is could be dangerous for the UAV. So a circle waypoint is created near here so that the UAV will be sent into a circle that is centered at R27 and then merge back to the flight path towards waypoint R28.

3. Results

The images taken by the imager system are orthorectified with the recorded georeferencing data. These images can be stitched together and form a mosaic of the river. The mosaic is projected onto Google map so that it is easier to see the difference between the actually river and our river map.

Fig. 7 The mosaic of orthorectified images

In the plot, we can see that the mosaic covers the entire river path that the UAV has navigated through. An outline of the river is displayed in the mosaic. However, the river path lacks coherency. Although on the whole the river path is recognizable, part of the river is disconnected or even missing.

Fig. 8 The mosaic is more successful when UAV is flying in straight line, less when it’s cornering

There are various reasons of this mapping error. Firstly, it is due to the sensor error. The IMU (inertial measurement unit) is the sensor which detects the pitch, roll and yaw values of the UAV. The reading from this sensor is estimated from the measurement of the UAV posture using a Kalman filter. Noise exists constantly in the measurements so that estimation error is inevitable, especially when the plane is turning or climbing dramatically. Secondly, the GPS data is not accurate enough. The GPS antenna on the UAV needs to be compact as well as low power consumption, its signal quality is hence compromised. During a flight the GPS error is as large as 5~10m, which is not an issue to navigate the plane, but for orthorectification, this error appear to be not acceptable. Thirdly, there are other issues such as the logging delay of the georeferencing data with respect to each picture; or the error from hardware set up, such as the cameras may not be mounted right, etc.

4. Conclusion

As a result, it is safe to conclude that by integrating DEM with the river flowline data, flight plan generation is remarkably easier, faster and more accurate. Especially the case when the mission requires UAV to fly above complicated terrains such as mountainous area.

To use the airborne images in GIS, more accurate georeferencing data is needed. Also a better feature based image stitching algorithm is required. Therefore the images can be better projected and imported to GIS as raster data.

Reference

Canon PowerShot SX100 IS: http://www.usa.canon.com/consumer/controller?act=ModelInfoAct&fcategoryid=144&modelid=15672

Gumstix onboard computer: http://www.gumstix.com/

Gphoto2 camera library for linux: http://www.gphoto.org/