UAS Field Day 2 and 3

Introduction: Aerial imaging via unmanned aerial systems (UAS) is becoming more and more affordable to the public. Simple digital cameras can be mounted on a kite or balloon, set on a timer to take pictures at intervals, and launched into the air. Of course, more sophisticated systems can be built such as the Y6 shown earlier, but complexity and expense are only obstacles if you want them to be. This exercise involved capturing aerial imagery via balloon and Y6 rotocopter. Each resulting set of photographs were then mosaicked into an orthophoto using Agisoft Photoscan. In normal aerial photos, there is error caused by the angle the camera is taking the image or natural changes in the surface that is being imaged. An orthophoto is an image that has been geometrically correct to remove any error caused by buildings, large changes in the surface, or if the images were taken at an oblique angle. The orthophoto is to scale with the real world objects and scenes photographed where a normal image could be stretched in certain areas. This is important if the images are to be used in any sort of mapping done needing a distance calculation. Agisoft Photoscan is a modeling software that allows images to be loaded and built into an orthophoto and 3D surface based on a generated point cloud. The newly created orthophotos were then georeferenced to the surface imaged with the use of ArcMap.

Study Area: Eau Claire Indoor Sport Complex and Soccer Fields, Eau Claire, Wisconsin

Methods:

Y6 Rotocopter: The Y6 rotocopter is the more sophisticated of the two platforms used to image the area around the indoor sports center. The systems has the necessary autopilot software and stabilization systems to carry a digital camera for the imaging. A flight plan was programmed into the autopilot software and Canon- SX260 digital camera was set to take a still image every 2 seconds for the duration of the flight. The resulting photoset held about 500 pictures. Of those 500, 105 were chosen based on clarity and continuity to be mosaicked in Agisoft Photoscan. The resulting orthophoto was then added to the ArcMap document with a satellite base map. Because the GPS on the camera logged in WGS-84, the orthophoto was projected using the “project raster” tool in ArcMap to UTM Zone 15N. The now projected orthophoto was georeferenced to the satellite base map using the “Georeferencing toolbar” in ArcMap. To reference the photo, click the “add control points” button from the Georeferencing toolbar. Then click an area on the orthophoto, such as an intersection, to set a ground control point. Now find and click on the same point on the satellite image. The orthophoto will be stretched to match the points to the points on the satellite image. It helps adjust the orthophoto transparency to help see the satellite image below. After the orthophoto aligns with the satellite image, select “rectify” from the georeferencing menu on the toolbar. This will save the edits made when the image was matched to the satellite base map.

    

Balloon: The balloon was flown at a height of a couple hundred feet. From the tether string, a camera mount was strung and held a Canon-Elph110HS digital camera. The camera was set to snap a still image every 2-3 seconds for 20 minutes. From the picture set, 96 were chosen to mosaic in Agisoft Photoscan. Before mosaicking the images, the GPX track needed to be appended to the images. 

Figure 1: GeoSetter offers a platform that makes linking the GPX track log to the images captured. Having a GPS location with the image improves the accuracy of the later orthophoto.

To assign the GPS location for each image, the 96 images were imported into GeoSetter (Figure 1). By clicking the “synchronize with GPS data file” button on the toolbar, GeoSetter aligned the images with their proper location and saved each image with a GPS coordinate. Now that the images have a GPS location, Agisoft Photoscan can properly make the mosaic and orthophoto. The orthophoto was then added to the ArcMap project with the satellite base map. The image was projected using the “project raster” tool in ArcMap from WGS-84 to UTM Zone 15N. Then using the same process as before, the image was georeferenced to the satellite image via the georeferencing toolbar. To save the referenced image, select “rectify” from the georeference menu on the toolbar.

Figure 2: Agisoft Photoscan uses the selected images added in the workspace and creates a 3D model of the surface. The images are linked, allowing the creation of a point cloud which the program uses to model the real world surface. 

Orthomosaic: Agisoft Photoscan was used to create the orthophotos from the collected aerial images. From the workflow tab in Photoscan, select “add photos”. This will bring the photos that will be mosaicked into the workspace. Now from the workflow tab, select “align photos”. The program will form a point cloud using all the images selected in this step. If a high number of images is used, the processing time will take longer. With both platforms using about 100 images each, processing took about 15 minutes. From the newly created point cloud, select “build mesh” from the workflow tab. This creates a TIN from the point cloud turning the 2D images into a 3D surface (Figure 2). Now to add the imagery on top of the TIN, select “create texture” from the workflow tab. Once the texture is added, export the orthophoto from the file tab in a TIFF format to be imported in ArcMap for georeferencing.

Georeferencing: Open a new ArcMap project and add a satellite image base map and set the map projection to UTM Zone 15N. Add the orthophoto TIFF created from Agisoft Photoscan. Before referencing, use the “project raster” tool to project the TIFF from WGS-84 to UTM Zone 15N. This will make the referencing process a little simpler and more accurate. Now to start referencing, open the Georeferencing toolbar from the customize tab in ArcMap. On the toolbar, make sure the TIFF is listed in the drop down box and select “add reference points”. Using the tool, click an area on the TIFF and match that spot to that of the satellite base map. Change the transparency of the TIFF to about 30% to help locate features such as intersections or corners of buildings to help accuracy. Repeat this process of clicking the image, clicking the satellite base map until the image matches the base map. Once the image matches, select “rectify” from the georeferencing drop down menu on the toolbar. This will save a new image that will now be georeferenced.

   

Results:

Figure 3: The georeferenced (bottom) and original orthophoto (top) comparison show how georeferencing can have a big impact on scale of an area imaged.

Y6 Rotocopter: The resulting orthophoto from Photoscan was clear and representative of the survey area. It took 15 ground control points when georeferencing to match the orthophoto to the satellite image (Figure 3). The edges experienced the most distortion especially noticeable in the residential areas.

Figure 4: Comparison of before and after georeferencing for the balloon imagery. The bottom image is georeferenced where the top is still disoriented in space over the map. 

Balloon: The orthophoto created from the balloon imagery was too distorted more on the edges of the image than the interior. The image was referenced using 8 ground control points on the images. Although it took less points to correct the balloon image, there was more discrepancies originally than in the Y6 orthophoto.

Figure 5: The Y6 has a compensation rig that allows the digital camera, while engaged in photography, to remain parallel with the ground. This keeps distortion in images to a minimum regardless of wind conditions and keeps the camera in a stable position. 

Discussion: The Y6 imagery was finer and more defined than that of the balloon. This may have to do with how the camera was mounted to the platform. The setup on the Y6 was an electric gyroscope that had automatic compensation for changes in the camera angle (Figure 5). If a gust of wind came up, the system automatically kept the camera parallel to the ground. The balloon’s camera was mounted using string and a couple brackets purchased at your local hardware store for under 5$. The rig kept the camera level but there was no way to compensate for wind gusts since it was attached directly to the string of the balloon. The Y6 camera also had the ability to automatically geotag each image collected during the flight. The balloon images had to later be processed using the GeoSetter software to link the GPX file captured on the rig to the images collected in the same location. If the balloon images were mosaicked without geotagging the images, the resulting orthophoto was displayed off the west coast of the Galapagos Islands in ArcMap, even after the correct projection had been defined. 

Figure 6: Without geotagging the balloon imagery before mosaicking, even with the defined projection, the orthophoto is displayed off the coast of the Galapagos Islands (red circle) instead of over Eau Claire, WI (green circle)

Even with the automatic tagging for the GPS locations, the cameras collect in WGS-84. It is important to remember this to correct the projections before you georeference it. The varying displays from different projections by the satellite base map could distort the accuracy of the referencing. Overall the Y6 performed more reliably but this exercise proves that even with a low budget, it’s possible to collect suitable aerial images to help map and survey locations.