Lab 10: Balloon Mapping Part II

Introduction

            With the knowledge gained from last week’s experimental launch of the balloon mapping rig, we set out on our final run at creating a high resolution mosaiced map of the UW- Eau Claire campus. Methods and discussion on how images were collected on April 15, 2013 and processed later that week to generate the final mosaic will be outlined by this post.

Methods/Discussion

            Again because of the short window of time that spanned our three-hour class period, a delegation of tasks was essential. Much like last week, class was broken up into small groups who were assigned to be responsible for a single element of the launch. Aside from logistical tasks such as measuring out 400 feet of sting that would tether the mapping rig to the ground and the transport of materials down to the staging area near the shed where the balloon would be filled with helium and attached to the mapping platform, several additional tasks were added.

            My team, consisting of Laurel H., Joey Q., Hannah B., and Zach W., set out with an array of handheld GPS units to record ground control points spread out over the entire AOI for the mosaic. GPS units used for ground control point collection were a Trimble Juno, Trimble Nomad, and Topcon. Although each team member was deployed with a different style GPS unit, the same ground control points, mostly light poles, fire hydrants, and other permanent features, were collected across all platforms. When a ground control point’s coordinates were taken, they were marked as waypoints within the GPS and given an accompanying attribute to signify which GPS unit mapped it.

           Following the collection of the points in the field, they were then converted into a point feature class within ArcMap. The purpose of collecting ground control points to later be used when georeferencing is to both assign a coordinate system to the unreferenced image, and to also allow for multiple images to seamlessly be mosaiced by positioning and placing them in their correct geographic location. Although ground control points can be established by visually interpreting corresponding points such as the corners of sidewalks or other permanent features between a referenced image and the unreferenced image, the addition of GPS generated ground control points would be especially useful for georeferencing imagery collected by the balloon mapping rig because much of the campus is under construction which would not allow images to be georeferenced to a previously remotely sensed base map. As it turns out, the accuracy of these ground control points did not allow for their use when later georeferencing the aerial imagery because of the variability among GPS units that sometime mapped waypoints up to a few meters away from their correct physical location.

Figure 1: Show above are two clusters of GPS ground control points. Each group of three were mapped in the same location, but notice how spread out they are. 

            Since we were unable to use the GPS generated ground control points to georeference our images, we were left to continue to brainstorm ways to generate highly accurate ground control points. Dr. Hupy came up with the idea of using an AutoCad coverage file showing building outlines. Although it was a good idea, the oblique nature of most of the imagery generated by the balloon mapping rig did not lend itself to be georeferenced to multiple previously determined ground control points, both GPS and AutoCad generated, because using those methods alone cased a highly distorted mosaic where building edges and other linear features did not match up.

Figure 2: Building AutoCad coverage overlain on a basemap of UW-Eau Claire

            Still wanting to avoid the user friendly, cartographically pleasing, but highly inaccurate freeware program Mapknitter, I set out to devise my own hybrid method to mosaic images together. I first found an on nadir image generated by the mapping rig that was in the center of my outlined section to be mosiced. Then using a set of guidelines I learned in a Remote Sensing (Geog 328) class offered here at UW- Eau Claire to visually interpret imagery, I established eight ground control points based on permanent features located on the ground. Once the first image was georeferenced, I then matched corresponding points from that image, to other over lapping images also generated by the balloon-mapping rig. This new method allowed for both a cartographically pleasing mosic, but also a correctly georeferenced one as well.

Figure 3: UW- Eau Claire campus broke up into sections to be mosaiced by each group

Figure 4: The section outlining Phillips Hall and south east corner of Campus to be mosaiced by  my team

            Once enough images had been mosiced together to cover the entire section outlined for our group, the mosic to new raster tool within ArcMap was run to create a raster mosic combining all of the input images. By carefully layering the images and placing them in the correct order within the mosic to new raster tool, I was able to avoid having the tether connecting the balloon to the ground visible on the final mosaic.

Figure 5: Final mosaic output for group 5 showing Phillips Hall,  and  surrounding south east corner of UW- Eau Claire

Conclusion

            Although our class was unable to create a perfect, seamless mosaic of the UW- Eau Claire campus, I feel that we were able to create one to the best of our ability with the resources available. Above that, the knowledge taken away about georeferencing and image mosaicing prove to be extremely valuable because it was gained through the process of group collaboration and troubleshooting which in my mind is the best way to learn.