Assignment #4 Point Clouds, Orthomosaics, and DSMs using Pix4D

Intro

This project allows us to go step by step thought data processing from the Wolfcreek flight.  This is the same flight we looked at in Assignment #3.  We start by using Pix4D, a photogrammetry and computer graphics software. Pix4D will give us a 3D map, 3D model or Ag Multispectral.  Since we are focusing on geospacial applications of unmanned aerial system (UAS), the data will be processed for a 3D map.  We will get back an orthomosaic  and digital surface model (DSM). To create map projections we need the raster files that are generated. Figure 1 shows them as the finished maps that we will create.

Figure 1: Orthomosaic & DSM without GCP correction 

Methods

To begin by launching the Pix4D application and create a new project.  This will be named by the date our mission was flown, location of the flight, the platform (aircraft) or more importantly the sensor, altitude that was flown, and coordinate correction system like GCP (ground control point) or ppk (post processing kinematics).  On the next page you will add images or give a directory to the folder with all of the images in it.  On the next page we selected the coordinate system we are using but since this is without GCPs nothing will change here.  Finally, we will chose what we want to do with our data.  We will be using the 3D Maps function on this project because we will get an orthomosaic and a digital surface model (DSM) output. Figure 2 shows a animation fly over of the orthomosaic in Pix4D.
                                             

Figure 2: Flyover Animation

Before the data processing begins, both number 2. Point  Cloud and Mesh and 3. DSM, Orthomosaic and Index.  Yes, these are the things that we actually want but we are going to run the initial processing first to check for anything that might be an issue with the data.  After initial processing is done we will be given the quality report where we can check for problems and see a preview of the orthomosaic and DSM (Figure 3). Part of what we are looking at is the number of images used and how many were rejected. If the number rejected is unreasonable, you may need to go in and manually create tie points in each image. Manually creating your tie points is incredibly time consuming but may help with processing issues.

Figure 3: Quality Report Summary Output

Now that we know the files will process, we can check the boxes for steps 2 and 3.  Then uncheck step 1 since it has already been done and press start.  The point cloud will be generated first.  The point cloud is a set of data points that span the surface of what ever it is that has been scanned or captured (Figure 4).

Figure 4: Pix4D Point cloud

Then the DSM, orthomosaic and index are developed.  These raster files are the same that we used in ArcGIS Pro but since we did not put in the GCPs in this set.  This DSM and orthomosaic are not the same as last labs.  We can see the changes when we compare the elevation values side-by-side in Figure 5.

Figure 5: DSM Elevation Comparison

Results and Discussion

Much like how we set up our folders for organization Pix4D does the same thing.  Each step, 1,2 and 3, has its own file folder and within each folder the appropriately related data organized in folders.  Pix4D will put the quality report in the initial folder in another report folder. With in the quality report we are able to find the processing times for each section. Figure 6 shows a table of the actions and the time it took to complete each section. 

Figure 6: Processing times

In the point cloud we can see voids or wholes that from the top down perspective are not visible.  It is from this exact top down that the photos were taken and there for the walls of some structure cannot be seen.  This lack of data leaves a void and if you try and triangle mesh them together, is also the reason that structures look melted.  This effect can be seen in one of my previous posts 3D Modeling and Orthomosaic Generation.  Some of these data processing failures could have also come from inappropriately identified points. When flying over surfaces that share very similar color values like large, flat, untextured surfaces or densely forested areas, photos can sometimes get confused and moved or removed. Tie points are not created between images due to their similarities or lack of data creating errors. 

Conclusion

Pix4D is important for our initial processing especially when developing a map because it creates our orthomosaic and DSM raster for us. Once these are created, we can pull them into ArcGIS Pro and do what we would like to add or change the already existing area.  With out it the process would slow down and leave most of GIS out of reach for most. The time requirements and entry level knowledge and skill set, not to mention all the time, would discourage and bottle neck the map making process.