Point Cloud, Orthomosaic, and DSM Creation using ArcPro

Introduction

What is photohrammetry?

Photogrammetry is utilized with mapping to help measure distances between objects and defining the location of a point in space.

What types of distortion does remotely sensed imagery have in its raw form?

Distortion occurs in remotely sensed imagery due to the fact that a two-dimensional image is being taken of a three-dimensional object. Objects seen directly from above will have no depth, whereas objects far from the center of the lens will appear to ben outward slightly. This problem becomes more noticeable the further away from the center of the image the object is.

What is orthorectification? What does it accomplish?

Orthorectification is a process that attempts to remove the negative effects (distortion as mentioned above) of an image. It can solve the issue of the tilting that happens in images to create a correct, more accurate representation of the space.

What is the Ortho Mapping Suite in ArcPro? How does it relate to UAS imagery?

The ArcPro Ortho Mapping Suite is a set of tools in ArcPro that can enable the creation of proper and accurate orthomosaics using orthorectification adjustments, among others. It is relevant for UAS imagery because all UAS imagery will have distortions that can be corrected due to the nature of the image.

What is Bundle Block Adjustment?

Bundle Block Adjustment is a process that attempts to stitch images together and reduce the amount of error of these images. After stitching the images together and reducing errors, a block of images has its precise ground location predicted, and is placed there. The block adjusted data can be seen below in Figure 1.

Figure 1: Flight data after block adjustment

What is the advantage of using this method? Is it perfect?

The advantage of Bundle Block Adjustment is that errors in the images and their projection can be minimized. Of course it is not perfect, and processing images with it can be very slow, but it helps give the most accurate orthomosaic currently possible.

Methods

What key characteristics should go into folder and file naming conventions?
Both the file and folder names should described what is contained in it. Folders should tell what type of data and what project is inside it, and files should let the user know what exactly it is.

Why is file management so key in working with UAS data
The sheer number of files associated with UAS data and imagery can be very overwhelming. Without proper file management, which includes proper folder and file names, data could easily be lost or forgotten.

What key forms of metadata should be associated with every UAS mission?
Every UAS mission should have the accompanying metadata: tools used to collect the data, specifications of the tools, other tool usage information (altitude), the specific coordinate system and projection being used, and the time and date the data was collected. All of these forms of metadata, and more, are shown in Figure 2.

Create a table that provides the key metadata for the data you are working with (Do not skip this).

Figure 2: Metadata for the flight data used in this lab

Results

Generate a table that shows the GCPs you used, and their coordinate locations.

Figure 3: GCPs used with the flight data

Generate Statistics for your DSM. Create a table that shows Max Elevation, Min Elevation, and Mean Elevation.

Figure 4: Statistics calculated for the DSM

Describe your maps in detail. Discuss their quality, and where you see issues in the maps.

The created maps show Dr. Hupy's house and property, and its location relative to the rest of Indiana. The three maps shown are the orthomosaic (Figure 5), the orthomosaic with the flight path and photograph points overlaid (Figure 6), and the DSM (Figure 7). Overall, the quality of the maps is very high. The unique characteristics of the land, such as the streets, driveway, houses, and trees that are separated from the rest are very clear. The shadows for these features are very accurate as well.

Figure 5: Orthomosaic Map

Figure 6: Orthomosaic Map with overlaid flight path and photograph points

Figure 7: DSM Map

Are there areas on the map where the data quality is poor or missing?

The DSM, shown in Figure 7, does not show the entirety of the created map, so some data is completely missing there. As for the orthomosaic maps, the large clusters of trees located on the west side of the maps has quite a bit of distortion and stitching issues. This can be seen much more clearly in Figure 8, where two completely different sections of trees that look nothing alike were forcibly stitched together. Issues such as the one seen below occurred all across the section of trees.

Figure 8: Stitching issues in the orthomosaic

Did the use of GCPs produce any noticeable changes?

The GCPs would have helped resolve some of the issues. Most of the data, especially in all of the non-forested sections was quite clear. However, the flight path did not focus on the forested section of the land, so there likely still would exist some of these stitching issues over that section of land.

How much time did it take to process the data?

During the creation of these maps and data sets, the lab was empty so the bandwidth to the server was freed up. Additionally, this is the cleaned up data set that contained 67 images, rather than the original 150+ images, so processing times on these maps is likely much shorter than it would have been otherwise. Processing times on various parts of the lab are shown below in Figure 9.

Figure 9: Processing Times

Conclusions

The orthomosaic tool in ArcPro can be a powerful tool, but the final product is very dependent on the inputted data. Utilized GCPs for the data would have helped to relieve some of the stitching issues noted above. However, the quality and type of sensor being used may be even more important. The orthomosaic tool in ArcPro did a great job with the flat terrain, individual stand-alone trees, roads, and houses. The tool did struggle to properly stitch the forest section of the map accurately, with many noticeable areas of improper stitching being shown.

The time investment with this data set was fairly minimal for the most part. As can be seen in Figure 9, the majority of the generated data took less than one minute to compute. However, three areas stand out as taking a significantly longer period of time than the rest. Generating the orthomosaic took over three minutes, which compared to the two longest sections is nothing. The second point cloud generation phase took nearly twelve minutes to complete, while the bundle block adjustment took just over twenty minutes. Given the quality of most of the map, it is remarkable that the whole process takes less than an hour of processing time. Using larger data sets with higher quality sensors and a large amount of ground control points would certainly spike the timing up into the multi-hour marks.

A higher resolution DTM from LiDAR data would make the final product more accurate because of the way LiDAR actually gathers data. There would be many more data points than currently exist with this data set. Elevation data could be gathered for each point specifically and even underneath trees and potentially between branches. The problems that exist with stitching could be severely limited using LiDAR data, but the processing time would skyrocket. This process might not work in more dynamic environments for the same reason there were issues with the sensor used for this lab. There is simply too much differentiating data, with potentially constant changes occurring, that could cause each piece of data to reflect a slightly different image and point in time. Dynamic environments are difficult to accurately capture, and many different sensors would struggle in a wide variety of circumstances.