Assignment 12: Processing UAS Data with GCPs PP

Introduction:

Like the previous assignment, this assignment involves processing imagery using Pix4D. However, this time the imagery will be processed with GCP's. Ground Control Points are an accurately surveyed coordinate location for a physical feature that can be identified on the ground.

Methods:

To begin open Pix4D, and choose New Project. Name the project specifying the date the imagery was collected, your last name, the type of drone used, and lastly add GCP to the end of it. The reason for to be so precise in the name, is to keep things organized and easier to find for later usage. In the edit camera model window, change the shutter model to Linear Rolling Shutter, and add the images labeled UWEC_Topcon in the GCP Manger window shown in Figure 1.

Figure 1. GCP/MTP Manager Window.

Double check the X and Y to make sure they are correct. X should be the Longitude/Eastings, and Y should be the Latitude/Nothings. When you hit OK the GCP's should appear in the flight area. The GCPs will apear as blue crosses, as you can see in Figure 2

Figure 2. GCP's on Imagery.

Once all the GCPs are on the flight area, and in the correct locations, you can start running the initial processing. To prevent issues with the GCP altitudes, and to assure the ray cloud editor works, use the Basic Editor option. Located the GCP marker in two photos for all 16 GCPs, and then reoptimize the project (see Figure 3).   Once the Green X matches the center of the marker you can process the final two steps.

Figure 3. GCP

When the process is completed the Raycloud should look like the one in Figure 4 below.

Figure 4. RayCloud view of imagery

Results/Discussion:

After bringing the data processed imagery into ArcMap, three maps were created. Just like the last assignment, those maps are a Orthomosaic model, a DSM model, and a DSM Hillshade effect, to get the best comparison possible.

Figure 5. Orthomosaic Maps.

For the Orthomoasic Maps in Figure 5, the map with GCPs is smaller and more accurate than the Map without GCPs. The cut out is the same, but as you can see the extent of the image for the map with GCPs is smaller than the extent of the map without GCPs.

Figure 6. DSM Maps.

As you can see in Figure 6, for the DSM map with GCPs, the Elevation is more precise and almost 100 meters higher in elevation than the DSM map without GCPs. THe less color variation means that the GCP map has a smaller range than the map without GCPs.

Figure 7. Hillshade Maps

Figure 7 shows the Hillshade Effect on the DSM imagery, and again you can see how the accuracy of the Z-field has caused a slight change in the elevation. Specifically, the GCP map has less drastic changes than the map without the GCPs.


Conclusion:

In conclusion, there is no doubt that GCP's help to make aerial imagery higher quality and more accurate. The maps with out the GCPs had with a Z-field that was over 100 meters off. In all three maps you can see that the extent of each map is less accurate when there are no GCPs. Outside of the extent change and elevation change, not much is different between the maps.

Sources:

https://support.pix4d.com/hc/en-us/community/posts/206071256-Altitude-Issue-When-Surveying-with-the-DJI-Inspire#gsc.tab=0

Assignment 11: UAS Data Processing NO GCPS

Introduction:

This assignment involves the construction of a point cloud data set, true orthomosaic, and digital surface model using Pix4D software. Pix4Dmapper is a software that creates several types of imagery, both 2D and 3D, by processing drone data. It does this through photogrammetry and computer vision algorithms. The assignment is separated into four parts. The first helps to get familiar with the product, and the second part involves actually using the software. Part three includes making some maps, and lastly part four is the written report.

Methods:

Part 1:

Look through the given Software Manual. Before starting the project, a high overlap is needed for Pix4D to process imagery. If the image acquisition plan isn't designed carefully, it may not have enough overlap. The plan depends on GSD of the project specifications and the terrain type to be reconstructed. The frontal overlap must be at least 75 percent, and the side overlap must be at least 60 percent.  If the user is flying over snow, sand, or uniform fields, their will be little visual content due to the large uniform areas. In turn there must be a frontal overlap of 85 percent and a side overlap of at least 70 percent.  The Rapid Check is an alternative initial processing system where accuracy is traded for speed. It processes faster in an effort to quickly determine whether sufficient coverage was obtained, but the result has fairly low accuracy. Pix4D can process images from multiple flights, however, results are best under the same conditions. The pilot also needs to maintain height flight. Pix4D can also process oblique images for mixed reconstruction, and can generate high quality point cloud data. GCPs are not necessary for Pix4D, however, they are highly recommended to most accurately represent elevation. When all images and data are processed, a quality report can be generated. It provides information on the accuracy and quality of the data collection. Any problems with the data can be found in this section.

Part 2:

To begin open the application, and choose Create a New Project. Make sure to include the Date, Site, Platform/sensor, and the altitude in the name of the project. Extract the data and imagery from Professor Hupy's shared folder into your individual folder. Then under the select images window, add all of the necessary images. Read the metadata, and notice the images are geotagged with a WGS 84 Decimal Degrees as the default UAS data. Change the Phantom 4 camera from a global shutter to a rolling camera. Also change the Camera Model Name to Linear Rolling Shutter. Leave the parameters as well as the Output Coordinate System. Choose 3D Maps for the Processing Options  Template, and start the process. You will then see the map view screen (Figure 1), go to the processing options (Figure 2).

Figure 1. Aerial View of Litchfield Mine. Each red
dot is a photo taken from the Phantom 4.

Figure 2. Processing Options Window

 Customize the processing options to  improve quality and efficiency. For better results, change the Raster DSM option to triangulation. Then wait for the Initial Processing, Point Cloud and Mesh, and DSM, Orthomosaic and Index steps to process. Once it is finished, create a quality report.

Figure 3. Shows the Summary, the Quality Check data as well as

the Orthomosaic and DSM before densification

As you can see in figure one, 197 out of 222 images were calibrated.

Figure 4. Shows the initial rayCloud image.

In the original rayCloud image, the red dots are the images that did not go through, this is due to tree cover.

Figure 5. Displays the rayCloud image with
Triangle Meshes and no cameras.

By taking of no cameras, and showing the triangle meshes, an accurate depiction of the actual landcover itself can be seen.

Figure 6. 

Results/Discussion:

Part 3:

Three maps were created to show display the Litchfield mine site including a Orthomoasic model, a DSM model, and a DSM Hillshade effect. All maps were created through the Phantom 4 imagery at an 80 meter elevation. The sensor was a 13.3 MP RGB Sensor, with a Rolling Shutter lens.

The first map created uses the Orthomosaic imagery. This map gives an great view of the different types of terrain. Although, it was not able to capture the images consisting of only trees, it provides very accurate detail of the different types of sandy and rocky terrains.

Figure 7. Displays a Orthomosaic model

of the Litchfield Mine.

The second map created uses the DSM imagery of site. The cool colors are the tallest elevated areas with dark blue being the maximum, and the warm colors are the lowest elevation with red being the minimum. As you can see the Northeast corner has the highest elevation.

Figure 8. Shows a Digital Surface Model
of the Litchfield Mine.

The third map created is Hillshade map derived from the DSM imagery. It does a great job of displaying the texture of the Litchfield mine. However, compared to the DSM imagery, it can be misleading for which areas actually have the highest elevation.

Figure 9. Shows a Hillshade Effect taken
from the DSM Imagery.

Conclusion:

In conclusion, all though the data process took awhile, Pix4D proved to be an exceptional program. Point cloud imagery is a great method for 3D modeling. The maps turned out great. The shortcomings were that only 88% of the images calibrated, but that is to be expected when it comes to tree cover. LiDAR would be the best way to get around this issue. On top of that their is no 3D GCP, meaning the Z field cannot be as accurate. Again the importance of Metadata is shown in this assignment to give viewers basic information on the maps, date, and drones. It would come even more in handy when processing multi-flights.

Sources:

https://support.pix4d.com/hc/en-us/sections/200591059-Manual#gsc.tab=0

https://support.pix4d.com/hc/en-us/community/posts/203318109-Rapid-Check#gsc.tab=0