Field Activity #7: UAS Flight Survey and Image Processing

Introduction

For this activity, we learned how to process UAS imagery by utilizing Pix4D software and use ArcGIS to preform the unsupervised classification technique. In this exercise, we processed Planet Imagery of the Chippewa Valley Area, UAS imagery from both drone flights from both Geospatial Field Method Section 1 and Section 2. These drone flights occurred in Putnam Park, a local park located behind Lower Campus, and next to residential areas.

This is the processed drone imagery of the November 6th drone flight of Putnam Park.

This is the processed drone imagery of the November 5th drone flight of Putnam Park.

Methods

Some important things to consider before flying a drone is wind direction and temperature. Our drone from the university would only operate 0-40⁰ C and with low wind speeds. Thankfully, although the temperature of the day of our flight was at freezing, we were able to get the drone up and flying.

After completing our drone fight on November 6th, Martin processed the 2D data, and 90 images of the flight, through Pix4D application. This application provided us with a lot of information of the drone flight, involving the elevation of the drone at each point, the type of drone, area covered, wind direction, how long the process was, etc. within the quality report.

This image displays the data being processed in Pix4D.

One error that occurred when processing the data after the drone flights, was although our imperial units in the field were set, they did not reciprocate within the Pix4D program, transforming the units into meters. What is also strange is that the meters to the amount of feet the drone flew did not match up. Therefore, this comes to show that processing drone data is a learning process and there is still so much to know.

This image displays the finished product after processing the data in Pix4D.

After processing different imagery (Planet Imagery of the Chippewa Valley (provided to us by our professor), UAS imagery from Nov 5 and Nov 6 Flights), we used that data in ArcMaps. I used the unsupervised classification using 4 classes, and for both the planet imagery and UAS imagery from the November 5^th drone flight, there turned out to be even a smaller number of classes. This is because depending upon the resolution and accuracy of the imagery, it directly impacts how many classes are produced.

Results

I then produced three maps of with the results of unsupervised classification. These maps were based on classifying the elements from the image. This was done by adding a new field (within the layer’s table), calling it “Class”, then enabled the editor tool to select the layer and type in what you interpret the elements to be. The three maps are shown below:

This map displays the USA Unclassified Image processing technique that was used on Planet Imagery, given to us by our professor, of Eau Claire county. The three classes that I have designated to this image involves Buildings and roadways, surface water, and vegetation.

This map displays the UAS Unclassified Image processing technique that was used with drone data from November 5^th, 2019. The three classes that I have designated to this photo involves build up, not classifiable (blank space), and vegetation.

This map displays the UAS Unclassified Image processing technique that was used with drone data from November 6^th, 2019. The classes I have designated for this data includes ground cover, not classifiable (blank pink space), snow cover, and vegetation/forest cover.

Discussion

With all data processed, it seems as if data from some flights are more descriptive than others, involving more classes and descriptive imagery. This makes me wonder how accurate or advanced drone technology is and why the data taken on November 6^th ended up with more accurate and higher resolution data than November 5^th’s drone flight data. Martin had explained how it takes a while to understand how to work with drones, so instead of thinking of error ridden data as a failure, it is good to think of it as a learning experience. In the long run, this is how you gain foundational knowledge about using drone technology and processing the data.

Conclusion

In conclusion, drone data and how to process it still has its uncertainties we have yet to understand, at least within UWEC’s Geography department. With this experiment however, it was very difficult to learn how to use a drone and learn how to process its data using Pix4D without specific guidance. This comes to show that when conducting a survey to gather data, it is best to understand your instruments fully in order to gather the best data possible in the field. Overall, it is important to look at these technological difficulties not as failures, but as adding a valuable tool to your toolbox.