Intro to Mosaicking

Introduction

This short report refers to a specific procedure taking part on the Balloon Mapping Project, where aerial images from the University of Wisconsin - Eau Claire (UWEC) campus were taken with a simple camera elevated by a helium balloon. Individual images are interesting but to see all of the images put together can give a much better view of the location.

It’s possible to do that in a number of ways, such as using MapKnitter, ERDAS Imagine and Arc Map. In a first hands-on the procedures, the software used didn’t matter much, but in this second activity, Arc Map was focused. The class worked together dividing tasks equally over the campus, and the result should be an update imagery for UWEC campus.

Methods

The very first step is to select the best images to be used. Since the camera can be tilt and not necessarily in the perfect focus, an analysis of the more vertical and clear is essential to have better results (Figure 1). Also, the images need to have an overlap between each other of at least 60% and cover the entire campus.

Figure 1 – Oblique Vs. Vertical Images

Although intuitively the next step seems to be just to put everything together and match, when you take simple pictures with a camera, they are not georreferenced, so a mosaic tool wouldn’t work at this point. Therefore, the Georreferencing tool in Arc Map is used to give the right coordinates to the points over the image (Figure 2). The accuracy is improved as much control points are added, so a minimum of nine points per image was established.

Figure 2 – Georreferencing

After the images are correctly georreferenced, it’s time to put them together. In this step, an important point is to figure out the order of the images: the best images should be on the top, and the worse on the bottom. Also, it’s necessary to try different ways to avoid the string between the balloon and the ground (Figure 3), working with images taken in different angles. Then, the Mosaic to New Raster tool is used to produce the mosaicked image.

Figure 3 – Presence of string in the images.

Since this process is time demanding, the 18 students in this class divided the tasks to increase the efficiency and quality of the results: if each students have less images to be georreferenced, it’s possible to do it more carefully and with a higher precision. Therefore, the campus was divided into six areas, where groups of three would work in (Figure 4). For our group, five images were georreferenced for each one, and then mosaicked.

Figure 4 – Campus divided in 6 evenly polygons.

Discussion

Precise ground points were collected to improve accuracy when georreferencing, however, all of them are located in lower campus, while the section taken by the group was in the upper campus. Therefore, both imagery and the buildings feature class could be used as a reference. The buildings feature class didn’t match with the imagery though, probably because of a distortion in the imagery (Figure 5).

Figure 5 – Buildings feature class Vs. Imagery

Although the best would be to stick with the most precise – the buildings – there were some areas where there were not enough building corners (Figure 6), so the imagery would have to be used. Using two sources as a reference that doesn’t match each other was not a good idea, so only the imagery was used.

Figure 6 – Area lacking in building to use as ground control points.

When the five images chosen by each component of the group were ready, they would be grouped as a layer to ease the use of transparency and maintain organization (Figure 7). In that way, it was possible to analyze both imagery and pictures at the same time.

Figure 7 – Transparency Settings

The control points were focused in the area where each component was responsible for, since the same area – not completely accurate – would be overlapped by a better georreferenced image. After the georreferencing were completed by all the components, the mosaick to new raster tool could be directly used from the JPG, without the need of exporting it as a raster dataset.

Conclusion

The georreferencing activity can be time-demanding and require attention to detail, which can make it a really hard procedure to be done for the entire campus. However, the division between all the classmates allowed this activity to be efficient and productive.

It was also an interesting activity since the class needed to use their own resources and talk to each other to learn how to use the tools and the theory behind each procedure. More in this section will be covered in the final report for the Balloon Mapping Activity, where each step for the entire project will be explained.