Processing Pix4D imagery with Ground Control Points

Introduction

    Ground control points (GCPs) are points of known location on the surface of the earth which are used to geo-reference imagery. To practice adding these points in Pix4D, this week I reprocessed the original Wolfpaving dataset with GCP data.

Methodology

     Since I had already worked with this dataset in the past, I simply went into Pix4D and reopened the project. The GCP coordinates were provided to me in the form of a text file (shown below).

Figure 6.1: GCP Coordinates in YXZ Format

These points can be imported under Project > GCP/MTP Manager. I had to specify whether the coordinates were written in XYZ format or YXZ. Usually Pix4D will catch it if you select the wrong one because the points will be very far away from the imagery collected. If it does not automatically detect this, the points can be viewed on a map and manually be corrected.
Once the GCPs have successfully imported, they need to be manually matched with the imagery. This can be done through the rayCloud Editor or the Basic Editor. I chose to use the rayCloud editor. Based on the field notes taken at the time of data collection, the GCPs must be located and marked at the appropriate position (Figure 6.2).

Figure 6.2: Finding and Correcting the GCPs

After this has been completed for every GCP, the data must be reoptimized under Process > Rematch and Optimize. This step aligns the blue and green circles (Figure 6.5). This will overwrite the data processed in step 1 which I why I encountered the below warning message. 

Figure 6.3: Reoptimization Warning Message

Figure 6.4: Quality Report Generated
after Reoptimization

Figure 6.5: Post Reoptimization the Green
and Blue Circles Match

Once this is complete, steps 2 (generate point cloud and mesh) and 3 (generate DSM, orthomosaic and index) may be run. The output is shown in Figure 6.6. The result of this step will be usable layers in ArcGIS.

Figure 6.6: Reoptimized DSM in Pix4D

   

Discussion

    Adding the GCPs to the dataset in this exercise improved the accuracy of the maps produced. Compared to Figure 4.7 of two weeks ago, adding GCPs aligned blue and green circles (Figure 6.5) which implies the that the ground from the data was successfully matched to the projection. The impact of this can be viewed in Figure 6.6 where the GCPs fall in the indicated spots.
    Once the full processing of the othomosaic, DSM and aspect was completed in Pix4D, I used ArcGIS Pro to point out the location of the GCPs (denoted in red on Figure 6.7). Further, I used insets to illustrate what the GCPs looked like on the ground as these were the marks I had to go through and point out in Figure 6.2.

    One thing I noticed when locating these points was that sometimes it was difficult to find them within an image or determine which one was actually used. Well-written field notes would help in situations such as the GCP pictured in the lower right of Figure 6.7. There are multiple arrows painted, and the data collector had to be consulted to determine which one should be used. Documentation would have easily cleared up this confusion without wasting much time.

Figure 6.7: ArcGIS Pro Wolfcreek Paving

Orthomosaic with GCPs

Conclusion

    GCPs are an incredibly useful tool for creating more accurate and higher quality datasets. Working with this data specifically made me appreciate using GCPs as well as field notes. Working with data without much context has improved me as a field operator because now I know to never skip out on the documentation. Such accuracy that I was able to achieve with this data set will be absolutely necessary for obtaining survey quality data or calculating volumetrics from the data in the future. 

Getting to Know the Living Atlas

Introduction

    As an emerging industry, UAS demands a lot of knowledge in a broad array of subjects. Learning GIS skills over the past few weeks has already expanded my skill set, and I am always looking to learn new things. I have found that the ArcGIS Living Atlas of the World is a great resource to consult. The library contains thousands upon thousands of crowdsourced layers to be used in projects. Many are free to use, however, some are considered “subscriber” or “premium” content. Searching this database is made easy by the filtering and sorting options provided. This week, I looked into a few different tutorials that interested me, I walked through the ArcGIS Living Atlas tutorial, and I made a few maps of my own.

Exploring ArcGIS Online

     While exploring ArcGIS online I came across a few tutorials that caught my eye. I think each of these has a valuable skill or lesson to be learned which I would like to add to my UAS skill set.

Estimate Storage Capacity with Drone Imagery

This tutorial walks through utilizing aerial imagery to create data products of a water storage basin and calculate its area. As this relates directly to UAS imagery, I think it will be good to review. I was surprised to see that there were not more ArcGIS Pro tutorials focused on UAS data. Maybe in the future there will be more. 

Figure 5.1: Estimate Storage Capacity with Drone Imagery

Figure 5.2: Estimate Storage Capacity with Drone Imagery

Lesson Plan

Georeference Imagery in ArcGIS Pro

This tutorial looks at aligning and transforming historical imagery that are missing spatial reference information. Understanding how to do this will give a better understanding of coordinate systems and how to find and add essential spatial reference information. Ideally this would be for lower accuracy operations where the exact location of something helps understanding but is not relied on.

Figure 5.3: Georeference Imagery in ArcGIS Pro

Figure 5.4: Georeference Imagery in ArcGIS Pro

Lesson Plan

Apply Subtypes and Domains to Vienna Hiking Trails

Here, you implement and apply attribute validation techniques to geodatabase features to ensure data integrity and simplify data collection and editing. This is an application I could see myself working on in the future. I approciate the attention to detail this tutorial adds to that hiking map. This specific application of updating and enhancing hiking trails stood out to me as someone who loves to hike.

Figure 5.5: Apply Subtypes and Domains to
Vienna Hiking Trails

Figure 5.6: Apply Subtypes and Domains to
Vienna Hiking Trails Lesson Plan

Fly Through South America in a 3D Animation
This tutorial animates a 3D tour of a famous geographer's epic journey. This would be an impressive skill to learn. Presenting materials in a creative and unique way is important to me which is why this tutorial stood out.

Figure 5.7: Fly Through South America in

a 3D Animation

Figure 5.8: Fly Through South America in

a 3D Animation Lesson Plan

Construct Realistic Buildings with Multipatch Editing
This tutorial shows how to create detailed 3D building models with real-world image textures. This could be incredibly useful in architecture, urban planning, inspection or real estate application. This stood out to me as a UAS project I would like to pursue in the future. 

Figure 5.9: Construct Realistic Buildings with
Multipatch Editing

Figure 5.10: Construct Realistic Buildings with
Multipatch Editing Lesson Plan

Living Atlas Tutorial

    This week I looked further into how ArcGIS Online works including the ArcGIS Living Atlas of the World. This included accessing Living Atlas content across the platform and discover the capabilities that these layers and maps can support through the tutorial here.

Figure 5.11: Get Started with ArcGIS Living Atlas

of the World

Figure 5.12: Get Started with ArcGIS Living Atlas

of the World Lesson Plan

    The tutorial takes you through as if you are preparing to study the potential impact of the Grand Ethiopian Renaissance Dam, which is being built on the Nile River in eastern Africa. To learn more about the current water levels in the region it lets us dive into the content available in the Living Atlas.

Figure 5.13: ArcGIS Living Atlas
of the World Homepage

    To look into the global soil moisture over time, I selected the average monthly soil moisture modeled globally (shown below). This is a layer put together by NASA and curated by Esri. It is important to select data that is relevant to the question at hand and created by a reputable source. 

Figure 5.14: GLDAS Soil Moisture

2000 - Present

    To get a better look at this I opened it in the map viewer. When I did this I noticed it offered me to be a Beta user for the new Map Viewer update - I will have to look in to this more later. 
    Through the Living Atlas, applications built off of the selected layers can also be viewed. These can be useful tools and can give more insight into the data! The "Apps" tab offers a list of top curated apps - this means they were specifically selected and approved by Esri! In this situation the Water Balance Application (Figure 5.15) would be great for consulting the soil moisture around the Grand Ethiopia Renaissance Dam (the area of study for the tutorial).

Figure 5.15: The Water Balance Application

Featuring the Data at the Grand Ethiopia

Renaissance Dam

    Contributions can be submitted for Esri curator review under the "Contribute" tab and can be tracked under the "My Contributions" tab. In order to qualify for review it must have at least a fully documented item detail page.

Utilizing the ArgGIS Living Atlas to Create my own Maps

    

Conclusion

    The ArcGIS Living Atlas is a great tool to consult when compiling maps. There is a surprising amount of local information that can be used for a variety of projects. I would definitely consider referencing this for future projects. It could save time, effort and money rather than recollecting the data.  

Processing Image Data in Pix4D

Introduction

    Pix4D is a suite of photogrammetry software designed specifically for UAS mapping. In this exercise I used it to make a "video animation trajectory creation" (aka: a fly through video) and generate GeoTIFF files which can be imported as layers in ArcGIS for further analysis. Made specifically for UAS applications, Pix4D is a great tool for processing and viewing UAS data and producing many of the deliverables clients are looking for.

Methodology

    This exercise is meant to serve as an introduction to processing UAS data in Pix4D. Sometimes this data may take a very long time to process which is why we only used a subset of the Wolfpaving UAS data from the previous post. As shown in Figure 4.1, when importing the data to Pix4D I make sure it displays the correct properties and is using the correct coordinate system as the metadata specifies. Usually this occurs automatically, but if for some reason it does not, the data set will be unable to be georeferenced.
    Figure 4.2 illustrates the flight pattern the UAS flew for the mission. In order to see if enough valid data was collected to produce accurate maps, Pix4D allows initial processing of the data (figure 4.3) which will then spit out some quality reports to be analyzed. One major setting that needs to be changed in this wizard is the camera shutter type. Many cameras including the one used in this project have a rolling shutter which moves over time and can distort the image. Pix4D can get rid of a lot of this distortion, but it must be specified prior to initial processing. The default is global shutter/fast readout which would be better to fly but can often be more expensive. Figure 4.4 shows the summary of this initial processing. One thing to note is that the average ground sampling distance is 2.35 cm which indicates that this data set will be very precise.  Figure 4.5 displays the overlap strength. From this map you can see what areas have high levels of overlap (center and along flight path) and which could use more data (edges). This tool may be able to help in situations where not enough data is collected and processing will take a long time. In that case, going out to get the data again may be more beneficial. The quality check verifies that that dataset will produce a good enough output to proceed. The result of this initial processing can be viewed in Figure 4.7.
    If the quality report checks out, the next step is to process a point cloud and mesh along with a DSM orthomosaic and index analysis. The first iteration of this can be seen in figure 4.8. Looking around at figure 4.8, however, there are many gaps seen in the data. In order to get a much cleaner image for figure 4.10, the  triangulation is calculated and added. This process interpolates what the missing space should be, but it is not always completely accurate and tends to underestimate the terrain.
   Once the terrain looked smooth and good it was time to make a video for the theoretical client! Using the "video animation trajectory creation" tool that Pix4D provides (figure 4.11), I was able to make a fly through of the area. Such videos may be valuable in certain circumstances where the land needs to be shown off for planning, real estate, or development purpose. The final product is shown in the section below.

 Figure 4.1: Importing the data to Pix4D and 

verifying it is correct.

Figure 4.2:  Viewing the flight path of the mission.

Figure 4.3: Preparing for the initial processing.

Figure 4.4: Quality report  - summary.

Figure 4.5: Quality report - overlap overview. 

Figure 4.6: Quality report - quality check. 

Figure 4.7: Initial processing with camera layer. 

Figure 4.8: Point cloud, mesh, DSM orthomosaic,

and index processing. 

Figure 4.9: Exploring the default DSM. 

Figure 4.10: DSM after triangulation applied.

Figure 4.11: Setting up the video animation trajectory
creation (fly through).

Results

   Taking a closer look at the terrain produced by Pix4D, especially during the fly through, the piles of raw materials looked very smooth and accurate shape-wise. In some locations it can look too smooth. Other than the piles though, smaller and skinnier objects did not fare well. There was a tractor originally pictured (lower right of figure 4.9), but in the final rendering you cannot even depict what was once there. This type of processing seems to be made more for constructing the landscape as a whole and less of individual, small objects. With the right, I think some  of this could be improved. 
    One of the big differences between this data processing and the previous week's exercise in ArcGIS is that the ArcGIS data utilized ground control points. Therefore, when the DSM in ArcGIS is overlaid on the terrain it appears at the proper height. When this happens with the Pix4D DSM it floats a bit because the altitude is not georeferenced to the ground points. 
   The fly through that Pix4D offers is a unique and fun deliverable. I enjoyed getting to use it but it was quite difficult to get the hang of. Despite having the technical knowledge to use it, it still requires a fair bit of skill. As seen in figure 4.12, the final result it mainly smooth save for a few transitions. 

Figure 4.12:  Wolfpaving fly through.

Conclusion

    Overall, Pix4D is a great tool for UAS data. It has some very useful features that are straightforward and easy to used. There are some things, like the fly through and the mission planning phase which are unique to this software suite. It does however, take a very long time to process the data compared to what I saw in ArcGIS. 

Building Maps with UAS Data

Introduction

    Possessing cartographic skills is essential for working with UAS data. Nearly every piece of data that the UAS collects can be plotted and portrayed on a map. Knowing how to do this in a clear and concise way is a marketable skill. Turning a drawing or aerial image into a functional map requires at least a directional reference, scale bar, locator map, and the source and metadata from the compiled data. Depending on the location and spatial patterns present, a number of techniques can be applied to create a functional map out of UAS data. In this exercise, the terrain elevation varied and there were multiple man-made developments present across the property. Hillshading, the process of adding light and dark areas to the map, was used to bring out a three dimensional view of the hills and structures. The objectives of this exercise were to apply cartographic fundamentals to create a map, and to set an example for a map I will create in the future.

Methodology

    The data was provided to me for this exercise by Dr. Hupy via the class server. It consisted of multiple raster files, including a digital surface model (DSM) and an orthomosaic (Figure 3.1), and metadata in a text file. A DSM is not to be confused with a DEM (digital elevation model). The difference is the DSM contains everything on the surface including trees, people, cars, etc. The DEM is edited more heavily to calculate the elevation of the surface without the noise on top. The relevant metadata for this data set can be viewed in the lower left corner of all provided maps (Figure 3.1-3.4). In order to manage the many files that are created when gathering and manipulating GIS data, it is important to develop sound naming conventions and keep an updated metadata file. In this case, the main folder I am accessing is named based on both the location and client’s business: “Wolf Paving”. Furthermore, all files within this folder include “wolfpaving” in the name as well, so they can easily be traced back to this folder. Metadata is essential to the dataset as well because it keeps track of the who, what, where, and when of the UAS mission.

    Moving on to the map setup; The basemap in this application is an elevation map. It displays ranges of elevation with simple lines. This gives the detailed map context without distracting from it. Before bringing the data into the map, I calculated the statistics on each data set. This gives important information such as the cell size, units, projection, highest and lowest elevations. Knowing this can reaffirm the data collected is accurate, the resolution of the data, and how to use it for further calculations. When loading the data into the map, I told ArcGIS to build Pyramids. This allows the frame to be recalculated faster when zooming in and out of the data.

    The raw DSM alone is dull. To make the features and elevation pop, a hillshade was generated from the data (Figure 3.2). The hillshade can be set to a color ramp of the user’s choice. The one in this map is a natural earth-tone scale as this is standard for representing elevation. In this case, it is set to be transparent over the shaded DSM for maximal feature distinction. While editing the map, the swipe tool can be used to strip back the hillshade and shaded DSM to reveal the orthomosaic for reference. This is helpful to quickly reference the surface structures. At this point the hillshade provides feature relief. This can be draped over the actual elevation to create a three dimensional figure. Depending on what the client requires, this may be helpful especially if there are features observed with varying heights. Another useful analytical tool in this scenario is the slope and aspect calculations. Slope colors the raster data to the steepness of each cell (Figure 3.3). Aspect can then be used to show the compass direction these slopes face (Figure 3.4). This can be useful for analyzing which areas will be exposed to the most sun and tend to be dry. In Figure 3.4 the final map illustrates the aspect as the top layer with the hillshade directly below for extra feature relief.  

Figure 3.1: Initial orthomosaic rendering of the

provided data.

Figure 3.2: A transparent hillshade analysis overlaid

on the DSM.

Figure 3.3: A transparent slope analysis overlaid

on the hillshade.

Figure 3.4: Final map portraying a transparent aspect
analysis overlaid on the hillshade.

Conclusion

    UAS data is a useful tool to both cartographers and GIS users because they can quickly, and accurately gather various types of data at a location for GIS analysis. With that being said, the data collected is not bulletproof. The cartographer still must have an understanding of the data type, how it was collected, and the limitations of it. If any of these points are ignored, the analysis and presentation of the data risks being inaccurate. UAS collected data can be a powerful tool for geographic analysis. It can stand alone or be supplemented with preexisting projections of the area collected from satellites, aircraft or other means.

Getting Acquainted with ArcGIS Pro

For this lab, I was following along with this ArcGIS tutorial and recapping what I learned.

...

Introduction to GIS

Geospatial data is data that has a reference to Earth. It must have a defined scale and a projection. Geographic information systems (GIS) differ from a standalone digital maps because a map is simply an image that portrays the earth. Often times, maps do not specify the type of projection portrayed, and sometimes they do not even have a scale. Understanding the importance of geospatial concepts and data is fundamental to working with UAS because in order to produce coherent and accurate data with UAS requires the user to be aware of the projection and scale of the images. Without these, little confidence can be put into the accuracy of the acquired data. This ArcGIS Pro both covered the basics of navigating ArcGIS Pro while also reinforcing some key geospatial concepts and fundamentals. Creating a map with multiple layers illustrated how datasets can be related to one another based on their projections. The workflow involved many datasets with different types of data that illustrated the full picture that GIS helps portray.

Methodology

Section 1: Explore the Study Area

    The tutorial starts out with the basics. I created the project, imported the data, and set the basemap. A basemap is quite literally, “the base of the map” or the background. This gives context for the data that will eventually be overlaid on top. For this scenario, since we will be looking at both roads and forestry, satellite imagery makes the most sense. There are many other options such as street maps and different color maps for maximal contrast with the data. From there, the locate tool can be used to locate the Brazilian state of Rondônia since all of the data will be based around this area. The datasets are then added over the basemap: Brazilian states, an outline of the Amazon ecoregion, and cities in the state of Rondônia. The symbology tab of the catalog pane is used to edit the visibility and appearance of each dataset. For example, the color and thickness of the roads and cities have been edited for improved visibility. All of these changes were then added to a customized Geodatabase for easy storage and reference.
    The next part took a deeper look at the projected road data. First, the Roads layer is added to the map. From there, the “Official roads” are selected via a logical expression built with the “Select Layer By Attribute” tool in the “Geoprocessing” pane. This creates a SQL expression which is then run under the hood to differentiate the official roads from the unofficial. The result of the search creates its own feature class which can then be colored to stand out. The result is shown in Figure 2.1.

Figure 2.1: Official Roads (shown in white) have been 

differentiated from unofficial roads (shown in purple).

Section 2: Compare Roads and Deforestation

    In order to analyze the impact the roads have on deforestation, another layer is added from the Rondônia database to illustrate the deforestation. To see the basemap through this layer of vector data, the opacity is edited under the “Effects” group of the “Appearance” tab. By turning the Roads layer’s visibility on and off, it is clear that the deforestation appears most frequently where the roads are built. In order to visualize the protected forests in the area, another layer is added and colored green to represent the protected forests of Rondônia which have very little overlap of deforestation (Figure 2.2).

Figure 2.2: Deforestation in Rondônia (shown in white)

versus protected forest (shown in green).

 Figure 2.3: Deforestation highlighted near existing roads

(shown in blue).

Figure 2.4: Percentage of deforested area near roads calculated

(shown highlighted).

Section 3: Predict the Impact of the Proposed Road

The proposed road raster dataset was located in the Rondonia geodatabase. In order to make vector data from the portrayed raster data, a “Feature” class is created using the same South America Albers Equal Area Conic system as the other data in this GIS. Once the feature class has been created, it was necessary to digitize the road via the “Snapping” tool and “Trace”. The symbology of this feature was then modified and the fields Name and Status were added to the “Attribute Table”. 
To find a total area of potential deforestation around the proposed road, the Planned Road layer had to be buffered to the same 5.5-kilometer distance and multiply the buffer area by the percentage of deforestation observed around existing roads. In this part, however, previously deforested areas were erased via the “Erase” tool so they were not included in the total. The result of this buffering is illustrated in Figure 2.5.

Figure 2.5: Proposed Road with buffer and predicted

deforestation (shown in Purple).

Section 4: Finish and Print the Map

    In order to present the data clearly and concisely on one map, the proposed road was edited to stand out and the basemap was muted, among other minor tweaks to the visuals. Once the visuals were pleasing, an inset map is created to give global context to the location of Rondônia. Figure 2.6 illustrates the final printed map.

Figure 2.6: Final result map.

Conclusion

Based off of my experience using ArcGIS Pro in this tutorial, it is a powerful tool for visualizing GIS. As with many complex editing softwares, the learning curve is steep but well worth the effort. The final result of the 4 hours of work I put into the tutorial was concise, clean, and easy to read. I could see ArcGIS Pro being a powerful tool for the UAS and even broader aviation world. There are certain applications that would benefit greatly from using it, given the right dataset and user. On the other hand, it could make some simple problems too complex. For example, if accuracy is not that important or if the sample area is too small, this software may take more time than is necessary. In the scenario with Rondônia, UAS could assist in making more accurate predictions of deforestation. When zooming in on the deforestation layer that was generated with satellite imagery, it did not look very accurate. UAS could produce clearer images in a shorter time frame.