Lab 5: Construction of a Point Cloud Data Set, True Orthomosaic, and Digital Surface Model using Pix4D Software

Introduction

         Pix4D is a premier software for the construction of point clouds. This program is user friendly and has a lot of intriguing features. This lab will consist of the use of Pix4D to process a set of unmanned aerial system (UAS) data points taken from South Middle School in Eau Claire, Wisconsin. This study area consists of the South Middle School garden (Figure 1). This lab is designed to accurately process the data set using Pix4D.

Figure 1. This map contains the city of Eau Claire, which is located in Eau Claire County, Wisconsin. The South
 Middle school garden is the location of the red box in the south-east portion of the map. 

To produce highly accuracy results, a high overlap between images is needed. Since, Pix4D uses similarities between images to match up locations, flat terrain with similar vegetation, such as agricultural fields can be difficult to analyze. It is suggest to increase the overlap to at least 85% frontal and 70% side for this type of flat terrain. A higher flying distance is also recommended. The final thing that can be done to accurately map agricultural land is to have an accurate image geolocation and use the Agricultural Template on Pix4D. Multiple flights can be used in Pix4D if it is ensured that the plan captures the images with enough overlap, that there is enough overlap between two image acquisition plans, and the flights are taken in as similar conditions as possible. Ground Control Points (GCPs) are needed when the data points have no geolocation. If there is no geolocation or GCPs the final result will have no scale, no orientation, and no absolute position information. Thus meaning that it cannot be used for measurements. After processing is done a quality report is created. This will flag any areas of concern and give an overall feedback on how the processing of the data went.

            Ground control points are important in the accurate representation of datasets, even ones with geolocations present. A GCP is a “characteristic point whose coordinates are known” (Pix4D Support). GCPs are found by using traditional methods, such an older map, or other methods, such as LiDAR. GCPs are used to georeference an image and/or “reduce the noise” in an image (Pix4D Support).

Methodology

            The data of South Middle school’s garden was taken by Joseph Hupy. The UAS data points then needed to be processed in Pix4D. The first step is to open a new project in Pix4D. The name of the file should be a good reflection of what it consists of so it will be easily found in the future. The next step is to add the images taken by the UAS. Once the images are imported into Pix4D the metadata will be available. The metadata will show if the data points are geotagged and the associated coordinate system. In this case, the data points being used in this lab are geotagged and have a coordinate system of WGS_1984_UTM_Zone_15N. The camera properties should also be looked at; default values are set but should be checked for accuracy. In this case, it was assumed that the camera mode was global shutter but it is actually linear rolling shutter. This was easily changed within the shutter mode but if not it found could have caused error within the processing of the image. When all the camera properties are correct, move on to the output coordinate system page. In most cases this information can be left with the default inputs. Next, a processing template can be chosen; in this case a 3D Map template was used. This is the last step and finish can then be selected.

            A map view of your image will then be projected onto Pix4D with the flight path marked with red circles. Before processing the data, uncheck step 2 and 3 of the processing. It is more useful to do the processing in two steps in case something is wrong with the set up and needs to be changed. Before starting the initial processing, select processing options. Various options within this can be selected to improve the quality of the output data. It was found that selecting triangulation in the raster DSM options produces a better output; therefore, this was done for this lab. After looking through the processing options, the start button can be selected to start the initial processing. After, the processing is done a quality report will pop up. It is a good idea to look over this information carefully before continuing with the rest of the processing. If there are any red flags, they will appear in this report. If everything looks in order within the quality report, the step 1 box should be unchecked and step 2 and 3 should be checked. Start can then be selected again to finish the processing of the data. Once this processing is done another quality report will appear (Figure 2). If everything went correctly the ray cloud data can be viewed. To view the map created turn off cameras and turn on triangle mesh (Figure 3 and Figure 4). After the processing is complete a folder will be created with the standard output data (Figure 5).

Figure 2. This image contains the final summary report from Pix4D after the full processing of the UAS data from South Middle school. 

Figure 3. This is a map view of the South Middle School garden after the processing of the data and triangle mesh was turned on. 

Figure 4. This is another image of the garden after triangle mesh was turned on. This view is looking from the southwest to the northeast. 

Figure 5. This image is the folder that Pix4D created after fully processing the data. 

Results

            The processing of the data went very well. 69 out of 69 data points were calibrated and used. The image created using Pix4D was very clear and appeared to be accurate. One concern from the quality report was the camera optimization. There was only 38% relative difference between initial and optimized internal camera parameters (Figure 6). This didn’t appear to affect the output images created but this does put into question the accuracy of the output data.

Figure 6. This is the quality check that was produced after the processing of the
data in Pix4D. The camera optimization with the red triangle is the area of concern. 

            After the output was created, the DSM was brought into ArcMap to create a map of the study area. The DSM was imported first and the color ramp changed to make the elevation change more apparent. A hillshade was created of the DSM to give the elevation more definition. A cartographically pleasing map was then created to accurately display the South Middle School garden (Figure 7).

Figure 7. Map of the South Middle school garden using ArcMap and the DSM created by Pix4D. 

            Overall, the software worked very well. It is surprising how user friendly it is. With little training and little instructions on how to processes the data; it was processed successfully. The processing of the data does take time and this data set was small. The processing of a large dataset would take quite some time to fully process, but Pix4D is user-friendly software that produces accurate images from UAS data sets.

Conclusion

            Pix4D processes UAS datasets so they can be accurately analyzed. The processing of the data points takes some time but is user friendly. The processing of the South Middle School garden went relatively smoothly. Once a new project was created the images simply needed to be inputted and the parameters checked and then processing could begin. The output on Pix4D is very clear and a very good visual. The ArcMap output is also a good representation but slightly less helpful due to the low relief in the study area. Overall, the use of Pix4D for the processing of UAS imagery works very well.

References

Data collected by Joseph Hupy.

Using GCPs. March 04, 2018, from https://support.pix4d.com/hc/en-us/articles/202558699-Using-GCPs