Adding GCPs to Pix4D software

Introduction

Last week we learned how Pix 4D is a powerful tool used in creating excellent orthomosaics and 3D models using geotagged images from the Canon SX260 camera and GEMs imagery with imported GPS coordinates. This week we are using our data collected during our fourth field activity and using ground control points or GCPs to increase the geographic and aesthetic elements of the created geo-referenced mosaics and models.

Summary of Methods

For data collection details refer to link above, but in review: Data was collected using a Canon SX260 camera with CDHK software installed so images could be taken at set intervals over the course of the mission. Before the mission began, 6 GCPs were placed around the study area and the high precision Topcon Positioning System was used to mark each GCPs latitude, longitude and altitude. The Topcon data was saved to a text file to be later imported to Pix4D. The SX260 was mounted to the Matrix UAV to take nadir imagery and a waypoint data was uploaded to the Matrix via Mission Planner. Data was then uploaded to a shared file to be analyzed.

There are multiple ways to add GCPs to a project in Pix 4D. Those being:

1. Measuring in the field with topographical equipment much like we did
2. Taking GCPs from currently existing geospatial data and finally taking GCPs from a web map service. So you don't necessarily need your own GCPs, however your final project will be a lot more accurate if you mark
3. Through collecting GCPs specific to your project.

There are three methods for tying images to GCPs. The first case is the one that we ran and requires that the GCPs have a known coordinate system. Here you can add GCPs to your project, run an initialization, mark the GCPs in the ray cloud and then complete the final two processing steps. The second method involves the a user who doesn't have geolocated images or images and GCPs are geolocated within a local coordinate system. In this case an initialization would be run first, followed by then adding and marking at least three GCPs with rayCloud and then adding more GCPs as necessary followed by running the last two processing steps. Finally, Pix4D allows you to, for any case mentioned in the previous two methods, to Add your GCPs Manually and mark them followed by doing the total processing all at once. This method requires the least amount of intervention, but probably won't produce optimized results in comparison to the first or second methods. More information can be found on the Pix4D support site, here.

Like mentioned earlier, our data was processed using the first method. First, our data which consisted of around 340 georeferenced pictures was added to our Pix4D project, following the same steps as last week. For this exercise we had to select a coordinate system, which for where our study was done was NAD83/UTM zone 15N, which can be found in the drop down menu. After creating the project, the text file containing our GCP information via clicking on the layers menu on the left column, selecting GCP and then Manual Tie Point Manager. It is important to know how your coordinate system is stored since Pix 4D displays them YXZ. Verify your coordinates are cooret before moving on. Once this was done, Initialization was started, making sure the other processing steps remained unchecked. This will show you your initial processing before you optimize the images to the GCPs.

Figure 1: Initial Processing Quality Report. Notice that no GCPs were used in the making of these
 mosaics as indicated by the exclamation point by the Geo-referencing check.  

Once Initial processing was complete, a quality report was generated and saved for comparison to the final processing. Now the GCPs were added. To do this, again the layers drop down menu on the left in rayCloud was clicked, followed by GCP and manual tie point manager. from here you can see each GCP you uploaded underneath. Click on the GCP you want to calibrate and on the left, a properties column will appear. Here you see your GCP that you clicked and how many images you have calibrated (should be 0 to start). Below that you will see an image similar to that of Figure 2. only you will not see the GCP marker. This is where the image believes the GCP is. To fix this, zoom out until you see the marker and then zoom back in, as close to center of the GCP as you can and click there. A yellow circle will appear indicating the new, calibrated location of the GCP.

Figure 2: GCP calibration images. Yellow circles indicate a calibrated image
and the Green X indicates the new, universal location of the GCP in your imagery

Continue to tag these images, marking the center of the GCP marker and do this for at least 10 images to solidify the location. As you move along, Pix4D automatically calibrates images which means you won't have to tag the hundreds of images that contain your GCP. Once you are satisfied with the calibration, click apply and your changes will be applied to your imagery. Do this step for each GCP and when you are done and changes have been applied, hit the optimize button which will optimize your imagery. Then when you are satisfied, click the process buton on the top menu bar and from that drop down menu, select reoptimize. Reoptimization will ensure your GCPs are utilized accurately throughout your project. Then a generate a quality report to see the changes.

Figure 3: Optimized data quality report. Notice how the study area grew in
size with the addition of GCPs.

From this point, the final two processing steps can be completed and depending on the processing power of your computer, it will take a fair amount of time.When processing is done, a map such as the one in figure 4 can be created. Plotted here are the GPS coordinates as they were recorded by several different means in a previous experiment with varying degrees of accuracy.

Figure 4: Map of study area with overlayed GPS coordinates as calculated by
various GPS enabled units

In ArcMap you can calculate the error of the various GPS systems using the measure tool in the top tool bar. In Figure 5, the distance between the most accurate GPS point and the iPhone geotagged photo is measured,

Figure 5: Measuring distance between the Topcon GPS unit and the iPhone 

Conclusion:

Pix4D remains to be a user friendly source for processing geospatial data and as shown above becomes even more accurate with the addition of GCPs. It is worthy to note that you do not need to have collected GCP data before collecing your imagery data, but do know that the quality of your project depends on the amount of work that you put into data collection, so for accurate results, collect GCPs with the most accurate means possible. Collecting GCP data is considered a must when it comes to conducting long-term surveys and I would uphold my suggestion from last week to use Pix4D for processing your data.