Aerial Imagery Mosaicing

Introduction
Our class has created and implemented a balloon mapping rig in the previous weeks.  This is an innovative and cost-effective way to collect aerial imagery.  In order to use the aerial imagery, the images collected by the balloon must be georeferenced and mosaiced.  This report concentrates on the process of both mosaicing and georeferencing images from our balloon mapping.

Methods
Two techniques were used to mosaic the aerial imagery.  For both techniques the images were uploaded to a desktop after the balloon was grounded.  The techniques used were very different; the first technique was using a website called mapknitter, the second technique was georeferencing through ArcMap.

Mapknitter is a website that provides tools necessary to "knit" together aerial images to create a map.  A Google Imagery base layer was offered by the website.  This base layer could be used as a reference for the placement and scale of the imported images.  The images had to be uploaded to the site one at a time.  Once an image was uploaded it could be scaled and moved so it could be located in the correct position.  This was done repeatedly with numerous images until the images were sufficiently placed.  Once this was complete, the map had to be exported so it was visible to all users of the mapknitter site. The map I created on mapknitter is seen in Figure 1.

Figure 1: Aerial Imagery map created with MapKnitter (mapknitter.org)

The second technique was georeferencing with ArcMap.  Because there was such a plethora of images collected by the balloon, the class divided sections of the campus into groups to lighten the workload for all students.

I started a new map document then loaded data that would be beneficial to the process of georeferencing.  The data included a polygon feature class for my groups section and a CAD polygon feature class of the buildings on campus (Figure 2).  I used an imagery basemap provided by ESRI in the beginning of the process as a reference.

Figure 2- Group Section & Buildings feature classes

After this data was added the process of georeferencing could begin.  This process if not very complicated, but it is time consuming and must be done carefully.  Below are the steps necessary for georeferencing.

1.) Turn on the georeferencing toolbox (Figure 3)

Figure 4: ArcGIS Desktop 10.1 Georeferencing toolbar

2.) Click the "Control Points Tool"

3.) Zoom to the current image

4.) Click a point on the current image that can be easily referenced to the buildings feature class

5.) Zoom to the Group Section layer

4.) Click on the area that matches the point previously selected on the current image

The image will move to a new location using the georeferenced point; do this repeatedly for each image until the image is placed in the correct area.  This process is repeated for each image until an area is accurately represented on the map.

Zooming between layers is helpful because the aerial imagery is not spatially referenced and is located very far from the needed area.

The georeference control points can be edited using the "Control Points Table" (Figure 5).  The control points are labeled by a number and when clicked on, the control point will be highlighted on the map.  Editing mainly consists of deleting control points if it distorts the image or the image's location on the map.

Figure 5: Control Points Table

Results

Figure 6: Final mosaiced raster of the aerial images

Discussion

Although I was apprehensive to use ArcMap to mosaic the aerial imagery together, I believe ArcMap 10.1 produced a better result and it was easier to georeference the images in this technique.  It was also a good resource to use because you could go back and edit the control points.  This wasn't available in Mapknitter, so if a project was greatly distorted you had to delete the image and start over.

The images taken from the balloon are not at a constant altitude or angle.  This makes it hard to match the images perfectly and some of the areas are distorted because of this.  To avoid this problem as much as possible, the images overlapped each other in many areas.  This problem can be seen in the area of the walking bridge and Schofield Hall in the final map.