Processing Data in Pix4D, No GCP's

Introduction

• What is Pix4D?
• Pix4D is a photogrammetry program that allows the user to generate 2D orthomosiacs, 3D surface models and point clouds with the geo-referenced imagery data they have collected. Pix4D allows for multiple different data inputs such as RGB images, thermal images and multispectral images. With Pix4D the user has the option to utilize a cloud based software and access their data and projects from multiple devices.  

• What Products does it generate?
• Pix4D generates photogrammetric models such as, but not limited to point clouds, orthomosiacs, surface and elevation models, textured 3D mesh and thermal maps. 

• Why is a so integral to UAS data processing?
• Pix4D allows the UAS user to generate products quickly and accurately from the data that was collected. Pix4D allows input from multiple common UAS sensors and does a good job at creating three dimensional models of the mission area. Pix4D does a good job at integrating flight planning and on the spot uploading (if desired) for UAS missions. The quick and accurate production of orthomosiacs and point clouds, as well as the ability inspect, annotate and adjust data makes for a very important tool in the UAS industry. 

Method

Part 1

• What is the overlap needed for Pix4D to process imagery?
• Recommended overlap is 75% frontal overlap and 60% side overlap, with respect to flight direction.  

• What if the user is flying over sand/snow, or uniform fields?
• When flying over special case areas increase the overlap to 85% frontal overlap and at least 70% side overlap, with respect to flight direction. With uniform fields increasing altitude may improve results. For sand and snow, adjust the exposure setting accordingly to increase contrast in the images. 

• What is Rapid Check?
• Rapid check is a tool to be used while in the field to ensure that the data collected covers the desired area and sufficient coverage was achieved. This is a low resolution production and is meant to be used while in the field and not as a final product. 

• Can Pix4D process multiple flights? What does the pilot need to maintain if so?
• Pix4D has the ability to process data from multiple flights but they specify the program will have issues with data sets over 2000 images. The user should split up the data set if it is over 2000 images. If the images are not geo-referenced the pilot must be using multiple ground control points to ensure accuracy. Pix4D has the ability to process data at different altitudes but there will be better results if the pilot maintains a consistent altitude. It is also important to maintain high overlap of each flight to ensure accuracy. 

• Can Pix4D process oblique images? What type of data do you need if so?
• Pix4D has the ability to process oblique images, the user should set a sensor angle and ensure there is high overlap between images. It is highly recommended when processing oblique images to use ground control points or manual tie points to ensure accuracy in the point cloud.  

• What is the difference between a global and linear rolling Shutter?
• A global shutter will capture the the entire image frame at nearly the exact same time while a rolling shutter will capture the frame line by line from top to bottom. The majority of cameras used in UAS operations have a rolling shutter and Pix4D will correct for the rolling shutter effect, it is important to specify the use of a rolling shutter sensor when processing the data. 

• Are GCPs necessary for Pix4D? When are they highly recommended?
• GCP’s are not necessary for Pix4D but they are recommended when processing large data sets, multiple data sets and for 3 dimensional models. 

• What is the quality report?
• The Quality Report is a report that the user can generate after each step of processing to ensure the data has been processed correctly and up to the standards of the user. 

Part 2

Move the data into a local drive on the computer to allow for faster processing time and ease of use. Avoid processing via a server drive. 

Part 3

Before we can begin processing we must add the desired data into Pix4D, for this project we are using the images captured from the Flight 1 data set. We can individually chose the images we want to upload or take the entire data set. Once the images are uploaded we need to select the correct coordinate system for initial processing, we are using WGS 84 UTM Zone 16 N, pictured in Figure 1 and 2 below. We then need to change the camera settings in Pix4D from a global shutter to linear rolling shutter, this will allow Pix4D to make the corrections for the rolling shutter effect. We will process the data using the 3D maps template. 

Figure 1: Initial processing details.

Figure 2: Initial processing details.

We will then run the Initial Processing step by it self. When the Initial Processing is complete access the generated Quality Report and ensure the data was processed correctly. Pictured in Figure 3 below is the summary from the quality report, the summary includes data such as the camera/ sensor model as well as the time for initial processing. Below the summary we can examine the data via the "Quality Check", as shown below only 66/67 images were used and there is an issue related to geo-referencing. There were no GCP's used in the processing and Pix4D is letting the user know of the possible processing issues encountered. We are also able to see previews of our orthomosiac and DSM.

Figure 3: Quality Report Summary.

Figure 4, below, shows the amount of overlap that was achieved in this flight, as shown in the figure the majority of the targeted area had good overlap with 5 or more images. There are issues in the south east corner of the data set with minimal overlap. Figure 5 and 6 show the flight path that was used for data acquisition, as seen in the flight path the grid does not extend out over the southeast corner, thus the image overlap is less than the 5 plus images. These images could have been the ones rejected due to outliers in the data set. 

Figure 4: Image overlap. 

Figure 5: Flight Path. 

Figure 6: Flight Path 2.

When initial processing is complete we can then run steps 2 and 3. Make sure to uncheck "initial processing" and then check "point cloud and mesh" and "DSM, orthomosiac and index".  The processing is now complete and we can analyze the models that have been generated. Figure 7 below shows the model without the triangle mesh and point clouds activated, while figure 8 shows the model with the triangle mesh and point clouds activated. We can see a clear difference in the rendering of the model, Figure 7 shows specific points in the data set and portrays a generalized idea of the data. When activating the triangle mesh and point clouds we are able to generate a more accurate model by tying together those points and the images we have captured.   

Figure 7: Model without point cloud and triangle mesh. 

Figure 8: Model with point cloud and triangle mesh. 

We then utilized the animation feature of Pix4D and were able to generate a video (posted below) that "flies" through the 3 dimensional model giving the user more perspective over the generated product. Figure 9 shows the path the animation takes to "fly" through the 3 dimensional model.

Figure 9: Animation Path. 

Maps

We can now create maps from the Pix4D orthomosiac and DSM by moving the data into ArcMap. Figure 10 below shows the Orthomosiac map, while Figure 11 shows the DSM map. 

Figure 10: Orthomosiac Map using ArcMap

Figure 11: DSM map using ArcMap

Figure 12: Meta Data from data acquisition

As seen in both maps pictured above there are slight deviations and misalignment from the orthmosiac and DSM when overlaid on the the satellite imagery base-map. There are multiple explanations for the misalignment, this can be due to errors within the satellite data or errors in the UAS data. Pictured above in Figure 3, the quality report warns the user of errors with regards to geo-referencing in the data set, from this information we can infer that there is an error within the UAS data. This error can be due to the use of Ground Control Points, as there were none used in the generation of this data.

The image quality within the orthomosiac is very good but we can see errors with dense vegetation in the trees to the western side of the map. There are significant elevation errors in the DSM as it shows there is a significant elevation change towards the center of the image. As mentioned before there was no use of GCP's in the generation of this data, the use of GCP's may fix the elevation problems in the DSM.

Conclusion

Pix4D allows the UAS user to generate accurate orthomosiacs and surface models quickly and easily while maintaining a high level of image detail. Pix4D allows for a close to end to end usage within the UAS field, the ability to process data collected on most UAS sensors and the ability to create and edit flight plans and paths. Pix4D also allows the UAS user to generate in-field products to ensure accurate data acquisition. The cloud option lets the user access their data on multiple devices and enhances the features in Pix4D.

The generated photogrammetry products are accurate and easy to analyze and alter when needed by the user. Total processing time is a drawback to using Pix4D as it takes a significant amount of time to process the data.