Volumetric Analysis with UAS Data

Introduction

In this lab, we learned about volumetrics in both Pix4D and ArcGIS Pro.  With these volumes, an analysis was done to identify the volume of the surface that has been modified by the removal or addition of surface material.  This is one of the most common uses of UAS in the construction industry, as companies need to know if they need to order more of the aggregate, or if their current inventory is sufficient.  This would be done prior to UAS technology with a rod and a surveyor in which the surveyor would have to collect coordinate data all around the pile using the rod and would even have to climb it to get this data.  Nowadays, the use of UAS greatly increases the efficiency of collecting data, as it cuts the time in half of what this process would normally take.

To add to our knowledge, in this lab, we used different methods of calculating the volume in Pix4D such as triangulated, fit plane, align with average altitude, etc., and compared the methods with one another to determine the most accurate methods when using the calculations of ArcGIS Pro as the base.  ArcGIS tends to be the most accurate way of calculating volume which is why it was used for the base.

Methods and Results:

The goal of this lab as mentioned in the introduction was to learn about the intricacies of volumetrics and volumetric analysis using both Pix4D and ArcGIS Pro.  There were two datasets used in this lab with the Wolfcreek Paving dataset with GCPs (used in previous labs) and the Litchfield dataset. Both of these were provided by Dr. Hupy.  The Wolfcreek Paving data was used for part 1 which entails calculating the volume of the area of interest in both Pix4D and ArcGIS Pro and comparing them. The Litchfield data was used for part 2 which entails the volumetric analysis, using the cut-fill tool to see the changes over time.  Both parts produce a final map that displays all of the crucial data.

For this lab, instead of a detailed methods section from the start, more of a workflow will be provided.  If interested in more details about starting a ArcGIS project like this, please click this link to review the previous lab in which this was provided.  For more information about the data used for Pix4D, please use this link.  The flowchart below in Figure 1: Flowchart provides a summary of the operations in this lab.

Figure 1: Flowchart

Part 1

For Part 1, since the dataset and DSM had been created in the previous labs, this project was opened up and saved to the new lab folder.  The cameras were turned off, and the triangle meshes were turned on. This can be seen in Figure 2: Cameras Off, Triangle Meshes On.

Figure 2: Cameras Off, Triangle Meshes On

The Volumes tab to the left was then selected.  This can be seen in Figure 3: Volumes Tab Selected.

Figure 3: Volumes Tab Selected

To calculate the volume of a stockpile, select “New Volume.”  This is the left button under “Objects” that has a cylinder and a plus symbol.  This can be seen in Figure 4: Create Volume.

Figure 4: Create Volume

This allows you to select the area of interest.  In this lab, the area of interest are the stockpiles in the upper right hand corner of the project.  This can be seen in Figure 5: Area of Interest.

Figure 5: Area of Interest

The area of interest can now be selected.  When selecting the area, ensure that there is enough of a base around the pile in order to compute the volume.  Below in Figure 6: Volume Selection Instructions, the instructions on how to do this can be found as well as the selected area for this lab can be seen in Figure 7: Selected Area. 

Figure 6: Volume Selection Instructions

 Figure 7: Selected Area

The volume of the area can now be computed, by selecting the “Compute” button.  This can be seen in Figure 8: Highlighted and Compute  under Volume 1.

Figure 8: Highlighted and Compute

The volume of the area has now been calculated.  This can be seen in Figure 9: Computed Volume.  

Figure 9: Computed Volume

This volume was calculated with the default setting of Triangulated which may not be the most accurate (this will be talked about later).  To change this, the settings icon can be selected. The Volume Settings can be seen in Figure 10: Volume Settings.

Figure 10: Volume Settings

In order to compare the volumes, each method was selected and calculated.  This can be seen below in Figures 11 - 16.

Figure 11: Volume 1 - Triangulated

Figure 12: Volume 2 - Fit Plane

Figure 13: Volume 3 - Align with Average Altitude

Figure 14: Volume 4 - Align with Lowest Point

Figure 15: Volume 5 - Align with Highest Point

Figure 16: Volume 6 - Custom Altitude

To better compare these volumes, a chart was created.  This can be seen below in Figure 17: Volumes of Different Methods. 

Figure 17: Volumes of Different Methods

The first part of part 1 is complete, and the focus can now be placed on ArcGIS Pro.  To start, like all ArcGIS projects, file management is key. Once again, if you would like to see how to do this, please click the link to review the previous lab for this information.  After the new project was created with the proper file management, the folder connections were added.  The DSM was added in which can be seen in Figure 18: DSM.

Figure 18: DSM

The next step involves creating a polygon feature class to extract the clip, sort of like cookie cutter.  To do this, expand the database and right click on the database created. Then click New and Feature Class.  This can be seen below in Figure 19: Create Feature Class. 

 Figure 19: Create Feature Class

The Create Feature Class menu appears.  For the name, I decided to name it Clip1.  For the type, “Polygon” was selected. This can be seen in Figure 20: Create Feature Class Page 1. 

Figure 20: Create Feature Class Page 1

On the next page, the fields are displayed.  A new text field was created. This can be seen in Figure 21: Text Field Added

Figure 21: Text Field Added

Next, the Coordinate Systems might need to be adjusted.  This is done on the Spatial Reference page. One of the best ways to do that is to import the coordinate system of the DSM.  This is what was done in this project and can be seen in Figure 22: Spatial Reference.

Figure 22: Spatial Reference

The rest of the settings can be left alone and the new feature class can be created.  After this, the feature class was dragged on the map. This will now be visible under the contents and catalog under the database which can be seen in Figure 23: Clip 1 in Database. 

Figure 23: Clip 1 in Database

Now, the feature class polygon can be drawn.  To do this, the Create button was clicked on under the Edit tab which can be seen in Figure 24: Create Button.

Figure 24: Create Button

  This brings up the Create Features dialog box which can be seen in Figure 25: Create Features.

Figure 25: Create Features

After ensuring the polygon feature is selected, the clip can be drawn out.  Similar to Pix4D, there should be room along the bottom of the pile. After the area has been drawn, the save button must be pressed to save the clip.  The symbology currently has a fill that covers the area, so this needs to be changed. To do this, right click on the clip under contents and select Symbology.  After this, the symbol image can be clicked on which pulls up the various outlines. For this, the black outline was selected which can be seen in Figure 26: Black Outline Selected.  

Figure 26: Black Outline Selected

The symbology has now been changed as seen in Figure 27: Symbology Change.

Figure 26: Symbology Change

This change can now be seen with the clipped area as seen in Figure 27: Clip 1 - Selected Area.

Figure 27: Clip 1 - Selected Area

This area now has to be extracted.  To do this, the Geoprocessing pane was opened and extract was searched.  This can be seen in Figure 28: Extract Searched.

Figure 28: Extract Searched

The Extract by Mask was selected.  This brings up the Extract by Mask pane where the input raster and feature mask data has to be selected.  After this, the output raster can be selected by saving where this would go. For this, the clipped pile must be saved to the database.  This can be seen below in Figure 29: Output Raster File Saved.

Figure 29: Output Raster File Saved

The clipped area is now it’s own layer as seen in Figure 30: Extract File.

Figure 30: Extract File

This area can now be viewed alone by unchecking the DSM under Contents.  This can be seen below in Figure 31: Pile 1 Clipped.

Figure 31: Pile 1 Clipped

Now, the volume of the area can be calculated.  To do so, Surface Volume was searched in the Geoprocessing pane which can be seen in Figure 32: Surface Volume Searched.

Figure 32: Surface Volume Searched

Similar to the Extract by Mask, the Input Surface has to be selected which will be Pile1_Clipped, the Output Text file has to be selected which will once again be saved to the database, and lastly, the Reference Plane has to be selected which will be Above the Plane.  The Plane Height has to be inputted now which was done by clicking 15 points along the base of the pile and recording them to an excel file to find the average. An example of this can be seen in Figure 33: Surface Plane 1 and Figure 34: Surface Plane 2.  

Figure 33: Surface Plane 1

Figure 34: Surface Plane 2  

The recordings and calculation in the excel file can be seen in Figure 35: Calculating Average Plane Height.

Figure 35: Calculating Average Plane Height

The standalone table can now be viewed under Contents which can be seen in Figure 36: Standalone Table.

Figure 36: Standalone Table

The volume can now be opened by right clicking and opening up the table.  This can be seen in Figure 37: Volume Opened and Figure 38: Table View.

Figure 37: Volume Opened 

Figure 38: Table View

This volume can now be added to the table that has the volumes calculated in Pix4D.  The table can be viewed below in Figure 39: Table of Volumes Calculated.  The results will be discussed in the discussion section.

Figure 39: Table of Volumes Calculated

The final map can now be created to better display all of the data.  This final map can be viewed below in Figure 40: Final Map of Part 1.

Figure 40: Final Map of Part 1

Part 2

Part 2 is similar to the second part of Part 1, however, it will use a different dataset, and a volumetric analysis will be performed.  The same steps will be used in this part as in Part 1, so the steps will not be as detailed in this part.

First, folder connections were added in order to open the DSMs. This dataset includes three flights which will be used for the analysis.  The three DSMs can be seen below in Figure 41: Opening DSMs.

Figure 41: Opening DSMs

These DSMs can now be dragged onto the map.  These layers added can be seen under Contents and in Figure 42: DSMs Opened.

Figure 42: DSMs Opened

Due to these being three separate flights (roughly 1 month apart from each other), there has not only been changes to the actual area, but the data cannot yet be compared due to the collection of the data varying slightly.  To explore these differences, the source of the DSMs can be view by right clicking and selecting Data Source. This can be seen in Figure 43: DSM Data Source.

 Figure 43: DSM Data Source

The variations of the DSMs can be compared in Figure 44: Raster Variation 1 and Figure 45: Raster Variation 2.  The first figure is the data collected 8/27, and the cell size differences can be seen when compared to the second figure that was collected 9/30.

Figure 44: Raster Variation 1

Figure 45: Raster Variation 2

A resampling needs to be done in order to accurately compare the site over time.  To resample the data, this was searched in the Geoprocessing pane. This can be seen in Figure 46: Resample Searched.

Figure 46: Resample Searched

After clicking Resample, the input raster has to be added in which can be seen in Figure 47: Resample Input.  For the output, the layer was saved to the database as well as named the layer with the resampling value of 10 cm which can be seen in Figure 48: Resample Output. This resampling value now has to be added in by entering 0.1 in both the X and Y value, and the resampling technique can be selected as Bilinear.  The final Resampling pane should look like Figure 49: Resample Settings.

Figure 47: Resample

Figure 48: Resample Output

Figure 49: Resample Settings

This resampling must be done for each of the three flights, so these steps should be followed for the other two layers.  After all of the resampling is done, these new layers should be visible under contents which can be seen in Figure 50: Resampled Layers.

Figure 50: Resampled Layers

A new feature class was then created to perform the clipping.  This feature class was named Clip2. After this, the pile can now be clipped.  The pile clipped can be seen below in Figure 51: Main Pile Clipped.

Figure 51: Main Pile Clipped

These piles need to then be extracted and must be done on each of the different layers.  This is where the feature class of Clip2 comes in handy, as the same area clipped on the first layer used can be used on the second and third layer.  The Extract by Mask feature is once again found when searching it in the Geoprocessing pane. After it is clicked on, the Extract by Mask Pane is displayed, and the first flight should match what is seen in Figure 52: Extract by Mask - 1st Flight while the second flight should match what can be seen in Figure 53: Extract by Mask - 2nd Flight.

Figure 52: Extract by Mask - 1st Flight

Figure 53: Extract by Mask - 2nd Flight

Once again, these extracted areas are now their own layers and can be toggled on and off.  The first flight area can be seen in Figure 54: Extracted Area with Overall Area, and the rest can be seen under contents in Figure 55: Flight Layers Extracted.

Figure 54: Extracted Area with Overall Area

Figure 55: Flight Layers Extracted

The symbology was changed to better show the elevations.  Each layer can be seen below with this reflected.

Figure 56: First Flight (7/22) DSM

Figure 57: Second Flight (8/27) DSM

Figure 58: Third Flight (9/30) DSM

To better show the elevations, I added a hillshade to each one.  The hillshade can be seen added to the first flight below in Figure 59: Hillshade of Flight One.

Figure 59: Hillshade of Flight One

The volumes were then calculated for each flight.  As a refresher, the inputs, outputs and reference plane can be seen below in Figure 60: Calculating Surface Volume.

Figure 60: Calculating Surface Volume

Once again, for the Plane Height, the average needs to be calculated for each layer.  This was done by finding 15 points around the stockpile and entering them on excel to find the average.  This can be seen for each layer in order in the figure below, Figure 61: Average Plane Heights Calculated.

Figure 61: Average Plane Heights Calculated

The volumes can be viewed in the Standalone Tables under contents.  This can be seen in Figure 62: Standalone Tables for Part 2.

Figure 62: Standalone Tables for Part 2

These volumes can be opened up by right clicking them and clicking Open.  These volume tables can be seen below for each layer in Figures 63 - 65.

Figure 63: Volume for First Flight (7/22)

Figure 64: Volume for Second Flight (8/27)

Figure 65: Volume for Third Flight (9/30)

For better clarity, I created a table with these volumes listed which can be seen in Figure 66: Volumes of All 3 Flights.

Figure 66: Volumes of All 3 Flights

The last step in this lab is to do a volumetric analysis.  This is done using the Cut-Fill tool that can be found by searching that in the Geoprocessing Pane.  This can be seen in Figure 67: Cut-Fill Searched.

Figure 67: Cut-Fill Searched

After selecting the first option, the Cut Fill pane is displayed.  The first flight is inputted for the before raster surface, and the last flight is inputted for the after raster surface.  For the Output Raster, the file was saved in the database and named accordingly. This can be seen below in Figure 68: Cut Fill Inputs and Outputs.

 

Figure 68: Cut Fill Inputs and Outputs

The Cut-Fill output is then displayed.  This shows the net gain, net loss, and unchanged over the course of the three months.  This can be seen below in Figure 69: Cut Fill Output.

Figure 69: Cut Fill Output

A final map was created to better display the data which can be seen in Figure 70: Final Map for Part 2.

Figure 70: Final Map for Part 2

Discussion

Based on the results for Part 1 in this lab, it was demonstrated that Pix4D has come a long ways with the accuracy of their volumetrics calculations. It was well known that Pix4D has not always been the most accurate tool when it comes to these calculations. This can be seen with the variation of each method. The biggest difference in these came when comparing the Align with Highest Point Method which calculated a volume of 7,439 m³ and the Aligh with Lowest Point which calculated a volume of 16,763 m³.  This was a difference of 9,324 m³ which is an extremely high variation.  When comparing the volumes with ArcGIS Pro which tends to be the much more accurate method to use, the Fit Plane method looked to be the most accurate with a difference of only 229 m³.  This would be the recommended method when using Pix4D to calculate this in the future.

Figure 39: Table of Volumes Calculated

The final map produced includes all of the data that was calculated in Part 1. This provides a quick overview that can then be presented to the company to show the amount of stockpile they have, and then they can determine if it is sufficient or not.

Figure 40: Final Map of Part 1

Based on the results for Part 2, it is evident that the stockpile has changed quite a bit over the three months. In the first month of July, it was seen to have a volume of 22,114 m³ that increased by a volume of 7,684 m³ in the following month of August, and then decreased by a volume of 21,209 m³ in the last month of September.

All of this would not have been able to be compared without resampling and accurate ground control points, as the slight change of collection and inaccuracy would produce a tremendous amount of error which would be pointless to even present to a company. This is critical to keep in mind in the future, as the data will not make sense without it. An example of how a well produced product that is accurate can be seen with this project below in Figure 70: Final Map for Part 2.

Figure 66: Volumes of All 3 Flights

Figure 70: Final Map for Part 2

Conclusion

This lab provided a real world example on how volumetrics are used, and how an analysis is done with this UAS data. In this particular situation, it was at a mining site, but many other environments can be used. For example, any kind of paving or concrete site would find this information extremely useful as well as any sort of national park that wants to examine the changes to a park overtime. With the advancement and use of UAS, this has greatly improved the efficiency and range that can be used in those kinds of environments, thus making this the best option. With this knowledge, it can inform the customer on an necessary actions that must be taken in order to continue smooth operations.