UAS Data Processing with Pix4D

Introduction:

Pix4D is an incredible program that allows the user to to convert aerial images taken by unmanned aerial surveillance vehicles into georeferenced 2D, and 3D surface models, point clouds, DEM, orthomosaics, and many more options.  It has many uses which include being used in agriculture, the mining industry, construction, and almost any other place as well. This program can create amazing models if done with the correct accuracy.

Pix4D Questions:

• What is the overlap needed for Pix4D to process imagery?

The recommended overlap for the user is at least 75% front lap and 60% sidelap.

• What if the user is flying over sand/snow, or uniform fields?

In an area with sand/snow 85% frontlap and 70% sidelap is needed.

• What is Rapid Check?

It is a quick processing that creates a visual surface of the area, but it will have a low resolution.  

• Can Pix4D process multiple flights? What does the pilot need to maintain if so?

Pix4D can process multiple flights and the pilot only needs to maintain the same vertical and horizontal coordinate system throughout the multiple flights. 

• Can Pix4D process oblique images? What type of data do you need if so?

Pix4D can process oblique images, but it is recommend to take images every 5-10 degrees.  It is also recommend to capture two sets of data at different heights. 

• Are GCPs necessary for Pix4D? When are they highly recommended?

They are not necessary for Pix4D, but they are highly recommend when there is no geolocation.  

• What is the quality report?

The Quality report produces a description with information informing the user how the initial processing looks. It gives a summary of all the checks of the data during the initial report and a low quality picture of the area.

Methods

To start creating surface models in Pix4D start with a new project named something that follows a naming scheme and save it in the correct folder.  Next, the select images screen will open which allows the images taken with the drone to be selected.  Find the correct folder containing the images and select them all to be processed.  Click next and and review the image properties.  With the drone used for our images the shutter mode needs to be changed to linear rolling.  Click next, to review the output coordinate system, click next to select the processing type.  For this case it will be "AG RGB."  To make the processing faster it may be useful to create an area of interest.  This is done by selecting map view and then selecting processing area.  Now click and draw a figure around the area intended and finish with a left click. Now to run only use the initial processing first to make sure the data is all good to go.  Finally, run the next two processes next to the initial processing to create the 3D surface model.

Calculate Surface Area

• Select View
• Select Ray Cloud
• Select New Surface
• Left click to make area and Right click to finish. 

Figure 1: Surface area. 

Measure the Length of a linear feature

• Select View
• Select Ray Cloud
• Select new PolyLine
• Click to select distance to be measure

Figure 2: Linear Distance

Calculate the Volume of a 3D object

• Select View
• Select Volume
• Click New Volume
• Draw out the points around the part of the image and right click to finish the shape. 

Figure 3: Volumetric calculation.

Create an Animation that 'flys' through the project

• Select View
• Select Ray Cloud
• click on the camera icon from the create box
• Choose generated waypoints or computer generated waypoints
• choose the duration and speed of the flight
• save the file

Results

Figure 1: DSM of the area of interest of the mine in Litchfield. 

Review

Pix4D seems to be a relatively easy program to use with everything layed out well on the main screen.  Even a beginner could pick up this program and upload the images and start making material with the help of other internet sources.  Its great that this program creates high quality data without having to have relatively high quality data to start with.  Although, this can be dangerous to some this is a wonderful program to work with.  It may not be as accurate and speedy as LiDAR, moreover it is still a wonderful tool if only aerial photo are available.  This program has many uses, it can be used in consulting for environmental firms, real estate, or even just taking aerial photographs as a hobby. overall this is a great program to use.

GPS Topographic Survey

Introduction:

This weeks lab was to use a dualff frequency GPS to get an accurate point that will be accurate to less than a meter.  This can compare to a smartphone GPS which will be around 10m accurate or a consumer GPS which is within 5m.  It also gives an elevation which can be used to create a continuous surface of the area. The GPS is connected through Bluetooth to a handheld device with the interface on it.  This makes it simple and easy to use in the field.  The important aspect for creating an effective surface area map is to get enough points to cover the areas with high relief and to create a study area that will accurately represent the area being studied.  The sampling method that was used is random sampling.  This means that random points were taken through the area rather than in one of the first labs using a sandbox when systematic sampling was used.

Study Area:

• 29 November 2016
• 40 degrees, Rainy
• Location: In front of Centennial Hall on the University of Wisconsin - Eau Claire campus.  The more detailed location would be tot he south of the circle area with the dancers statue.  

Methods:

Materials

Topcon Tesla Field Controller 

Topcon Hiper SR Positioning Unit

Tripod Stand (With level)

Figure 1: Dual band GPS in the background, and Dr. J Hupy holding the Topcon Tesla Field Controller to collect the first elevation point. 

The Data was collected using a dual frequency GPS which included using the materials above.  The information was Bluetoothed to a handheld which was downloaded to a computer.  Dr. Hupy processed the data and shared a text file.  The text file was then processed in Excel to be able to upload it into ArcMap using the import XY tool.  Then, the data can be used to create continuous surface raster layers using the IDW, Kriging, Natural Neighbor, Spline, and TIN tools.


Results/Discussion

Figure 2: Spline Interpolation of Elevation in a 2D view

Figure 3: IDW Interpolation 

Figure 4: Kriging Interpolation Method

Figure 5: Natural Neighbor interpolation 

Figure 6: TIN interpolation

Figure 7: Spline 3-D Interpolation

Figure 8 IDW 3D Interpolation 

Figure 9 Krigling 3D Interpolation

Figure 10: Natural Neighbor 3D Interpolation

Figure 11 TIN 3D Interpolation

Out of all of the different interpolation methods the most accurate one would be Natural Neighbor.  It really does the best at accurately representing the hill and the amount of relief.  The Krigling interpolation also did a decent job at representing the hill, but it did not do as well of a job at getting the amount of slope on the hill. The IDW interpolation method seemed to have too many holes or low spots in areas that were not actually lower, and the spline interpolation method was wavy and did not really do well with the edge of the slope on the hill.  The TIN did create a good image, but seemed to be edgy.  The 3D version of the IDW seemed to have some weird hills created at the top and some holes at the bottom.  This created a weird landscape that does not represent the surface well.

Conclusion

For making a continuous surface map it is important to collect enough points that would make the interpolation more accurate.  It is also important to accurately set up the GPS Unit, and it is important to have it level and upright.  This will create the most accurate data points.  It is also important to collect enough data points in the region with higher relief to represent the real world.  Using a dual band GPS also will create a more accurate point, but it does take longer than a cellphone or a consumer grade GPS.