Processing Data in Pix4D with GCPs

Introduction

In this lab I integrate GCPs into the Pix4D process. This lab is very similar to the previous with the main difference being the addition of GCPs along with the extra steps that entails.

GCPs or Ground Control Points are points of known coordinates in the area of interest. GCPs are essentially required when the absolute accuracy (position of the area of interest on the earth) of a project is important although this requirement can be waived if the images have geolocation. Images with geolocation utilize GPS data to record the position of an image when it is taken. Due to the innate inaccuracy of GPS, additional use of GCPs along with image geolocation can reduce the shift of the area from meters to centimeters. GCPs have an absolute minimum number requirement of 3 to be taken into account with each being present in at least 2 images. The recommended number is between 5-10 with additional GCPs having diminishing returns, even in large projects. If there are more GCPs than necessary, then you can convert some of them into check points. Check points can be used to assess the absolute accuracy of the data when your have already hit a point in GCPs that they aren't very helpful anymore.

Methods
The steps for this lab are similar to the previous one.

Stating a New Project
To start, create a new project and add in the images directory. In image properties don't forget to set the proper shutter model as seen in Figure 1.

Figure 1: Linear Shutter

Also, make sure that your output images coordinate system matches the coordinate system being used by the GCPs. The GCP coordinate system used should be found in the meta data.

Adding the GCPs

To add the GCPs go to Project in the top left of the window and click GCP/MTP Manager (shown in figure 2.)

Figure 2: GCP/MTP Manager

Verify again that the coordinate system is correct. To add GCPs click on Import GCPs and then click on browse in the popup window. From there you will add in the GCP file (often a text document) and check that the correct coordinate layout is being used (XYZ or YXZ.) A simple way to check this is by knowing the area's general coordinates and referencing this knowledge when reading the GCP file. Figure 3 shows the window where one selects the coordinate order as well as the browse button for selecting the GCP file.

Figure 3: Coordinate Order

Once you add them, you should be able to see the GCPs on the display. Now you can begin the initial processing.

Once the initial processing is finished you may notice that the GCPs are not exactly on the ground (Figure 4). This is a problem with DJI thinking that it's in a different datum but it can be fixed.

Figure 4: Floating GCPs?

To start fixing the GCPs select one of them by clicking on them (easier to do from below so you don't accidentally click on something else.) Once a GCP is selected it will show what images have that GCP in them via a line. On the right side of the window you will see information regarding the selected GCP as well as the images that contain it. In the images you may notice that the marked location of the GCP is not quite lined up with it's actual location (in this case marked by a chevron.) By clicking on the GCP's actual location in the image and then clicking apply, you can shift the position of the GCP in Pix4D to be where it actually is. Repeating this process with several images, and in all of the GCPs, should move them to their proper positions. In Figure 5 you can see that each GCP move a bit both vertically and horizontally.

Figure 5: GCPs shifting

After you have finished setting the GCPs go to the top left of the window and, under Process, select Reoptimize. (Figure 6)

Figure 6: Reoptimize

Once you have reoptimized you will see that both the image camera positions and the GCP positions have shifted. (Figure 7)

Figure 7: Shift

Now that the reoptimization is finished, you can begin the 2nd and 3rd part of processing, the "Point Cloud and Mesh" and "DSM, Orthomosaic and Index." Figure 8 shows what the triangle mesh looks like after the processing. 

Figure 8: Triangle Mesh

Figure 9 is the triangle mesh with the point cloud on and the camera positions off.

Figure 9: No cameras

Maps

The maps here are supposedly more accurate than the ones from the previous lab although, I confess, I don't see much of a difference. The last assignment was of a different location so there is no way to overlay the data sets to compare them. Figure 10 is the map I made from the orthomosaic and Figure 11 is the map that I made from the generated DSM.

Figure 10: Orthomosaic Map

Figure 11: DSM Map

Conclusion:

In conclusion, GCPs are essential for maintaining the absolute accuracy of your data as long as they are gathered in the correct way. GCPs should be well marked on the ground and the metadata should indicate exactly where on the marking the GCP was created as well as a general description of the the area that it was placed, this information should also be included in any field notes that you take. Check points are very similar to GCPs in that they are marked at a very specific, known, location. Check points are used to assess the absolute accuracy of the model where GCPs are used to improve it.