Pix4D Imagery with GCPs

Introduction

For this lab we will use the Pix4D processing software to process data taken from a phantom drone at Litchfield. We processed this data last week but this time we will process the data using Ground Control Points. This will yield highly accurate Pix4D maps that will create DSM maps of higher quality.

Methods

First you will need to open up Pix4D and at the images you wish to process. Once you have the images loaded into Pix4D you may add in GCPs using the GCP manager.

After you add in your GCPs you need to make sure that your Y coordinates are Northings and your X coordinates are the eastings. You may then do the initial processing of the data. After the initial processing is completed you will notice that the GCPs are not tied down to the ground.

We will now go through the process of placing the GCPs in the correct locations. First go to the Map and click on GCPs until you find point 0. When you click on a GCP it will bring up the GCP manager.

We will then select basic editor to add the correct locations for GCPs. Once you open the basic editor you will go through a few images, 2-5, where you can clearly see the GCPs.

You will point your cursor on the GCP in the image to tie it to this location. You will do this for 4 or 5 different GCPs and then re optimize your image.

After you do this for about 5 GCPs you can reoptimize the image This will allow you to more easily tie down the GCPs. You will have a group of images pop on the left hand side once you select a GCP from the layers column. From here you can correctly place the target on the GCP. You will notice some are close to the GCP but do not directly fall on the target. Once you do two or three you can select automatic marking and it will do the rest of the images for you. You will do this for all the GCPs that fall within the specific flight log.

After you complete this step all the GCPs should be tied to the ground and you can begin the steps 2 and 3 of processing.

After you finish steps two and three of processing you will end up with a raycloud map that has GCPs tied down.

Once you have both projects processed and tied down. You can create a new project that will merge both projects into one. Since the GCPs are all tied down this processing should complete slightly faster. If everything is tied down correctly the maps should perfectly overlay each other and GCPs should be tied down correctly.

Results

The merged DSM file within ArcMap should turn out pretty good if all your GCPs are tied down correctly. The final project will look something similar to figure 1. The map turns out a lot more accurate when compared to the basemap. There are some locations where the data on the maps is skewed. This would be in the south east corner of the map. This is the area of vegetation from the forest so this makes sense as to why the DSM file looks weird here.You notice the highest elevation within the map are areas to the northeast and the large piles of material through throughout the mapping area. The lake seems to be the area of lowest elevation which also makes sense. The GCPs allow us to easily connect separate flights into one data that overlay each other correctly. The Pix4D software seems to be very user friendly. It may be time consuming but easy to use. If you tie down the GCPs correctly you should run into little problems.

Figure 1

ArcGIS Pro Tutorial

Introduction

This lab will follow a tutorial from at arcGIS online here. The tutorial will teach us how to process UAS data and classify the surface images based on certain specifications

Methods

First click on the link and open the demo map provided in the tutorial. You will be directed to open the Calculate Surface imperviousness task.This will allow you to extract three bands to distinguish surfaces. They are displayed on the fly so no new dataset is created.
Follow the instruction 1-4 and click create new layer.

This will yield a map like figure 1

Figure 1

This creates an image that is easier to read. The vegetation shows up in the infared. This will make it easier to distinguish features in step two.

In this step we will segment the image.

First select Group Similar pixels into segments in the task bar.
Then complete steps 1-4 to create the new layer. This will create an image that looks like figure 2.

Figure 4

You will proceed to step 2 where you can review the segmentation before you move to the final step.This will complete lesson one.

Lesson 2
Open up the classify imagery task and open up arc map. In arcMap add both the Louisville Neighborhood and Segmented images. You will process classify this data in arcMap using the image classification toolbar. Next draw rectangles over a few houses in the north west corner.

Figure 3. Squares draw using image classification toolbar. 

Next select all of the class and merge them giving them the new class name grey roofs. You will do this with all the feature until you have a map that looks like figure 4. You can then save the classifications.

Figure 4

Next run the Train the Classifier task in ArcGISpro. And complete steps 1-5.

Next run the classify imagery tool following these steps yield a map like figure 5.

Figure 5

You will then run step two of the task to reclassify the data. This yields a map like figure 6.

Figure 6

In lesson 3 we will create Accuracy assessment points. You will use the create accuracy assessment points tool to get figure 7 with 100 accuracy points.

Figure 7

In the final project you will tabulate the area and clean up the attribute tables one the tables are joined. This will show you parcels of land that have the highest impervious surface. You will end with a map that looks like figure 8.

Figure 8

 Conclusion
This data produced a map that shows areas that have the highest area of impervious surfaces in darker colors. This happens to be areas of roads. This can help city planners determine storm water fees based on areas of lower impervious surfaces. If we have our own similarly accurate data we can go about the same process to classify the image using our own specifications.

Pix4D Processing Software

Introduction

Pix4D is an imagery processing software that is able to process data collected from various UAS systems. It is currently the best processing software for constructing point clouds. The program is able to process the data so it can be used in Geospatial programs like ArcMap. The program is easy to use and their online manual is easy to use. The processing part is time consuming but the speed can be adjusted dependent on the output quality required.

Preliminary Questions?

Look at Step 1 (before starting a project). What is the overlap needed for Pix4D to process imagery?

The recommended overlap is 75% frontal overlap (flying direction) and 60% side overlap (between flying tracks).

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

                Sand/Snow require a higher overlap of 85% frontal and 70% side overlaps. They also require the exposure to be set to get as much contrast as possible. Uniform fields require same overlap and that the UAS is flown at a higher altitude.

What is Rapid Check?

                It is an option for quickly processing data but in turn lose quality. It is a quick way to check data while in the field.

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

                The program can process many flights if the flights were maintain at the same altitude.

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

                Yes as long the images have angles recorded for the data.

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

                You do not need to place GCPs but the lack of GCPs will create less accurate data. It is highly recommended to use GCPs to reduce the amount of possible vertical and horizontal inaccuracies.

What is the quality report?

                The quality report gives a summary of all the data processed. It will allow you to verify the accuracy of the data before you start to use it. It shows you how the data was processed and parameters to the data. It also can show you areas where the data needs to be fixed or recollected.

Methods

For this lab we were giving two sets of separate flight logs. We were then instructed to process this data in Pix4D. You first open up PIX4D and create a new project.

It is important to name the project accurately. Include the date, site, platform, sensor, and altitude. Then add all the image flies from one of the flight logs. We will do the second flight log separately. Here you can look at the data points and edit the camera settings so they are correct for your camera.

Then select process data with 3D maps.

 Then click ok. You will now do the initial processing. Make sure to uncheck point cloud and DSM options. This will slow down the whole process if not unchecked.

You can edit the processing options to choose speed and quality of the output data. Once the initial processing is completed it will give you a quality report.

The quality report is very useful because it will tell you the accuracy of the data and whether you should proceed to steps 2 and 3 of processing. Once steps 2 and 3 are completed you can turn of cameras and turn triangles on to view the map created from the processed data.

There are some areas where the overlap is poor. These are areas at the edge of the map where the UAS made fewer passes. The middle of the map the UAS goes back and forth creating plenty of overlap.

Video Animation of the Data
https://www.youtube.com/watch?v=4-CIyOPjLMs&feature=youtu.be

Conclusion

Litchfield Map 1

Figure 1 Litchfield Flight 1

Figure 2 Litchfield Flight 2

Conclusion
Figures 1 and 2 show the DSM data collected from the Phantom 3 processed with Pix4D. There is some distortion in the corners where there was lack of overlap.
Overall the program is easy to use. The data processing is time consuming but as long as the pilot collected good data, the processing yields great results. The data can then be imported to ArcMap to further be analyzed as in figures 1 and 2.