Pix4D: Processing UAS Data

INTRODUCTION

Unmanned aerial surveillance photography is an efficient, affordable method in gathering data for small to medium-sized projects for a wide variety of uses. These platforms have a high-degree of customization with various attachments to achieve their goals. Besides controlling the flight and collecting data, the data processing can be an intensive process that involves careful attention to produce accurate results. Pix4D mapper is the most common software in UAS point cloud processing and is demonstrated in this report. Data was collected in the previous report UAS Platforms and GPS Units for Ground Control at the Litchfield mine located southwest of Eau Claire, WI. 

METHODS

Before processing UAS data in Pix4D, a basic understanding of the software and goal-specific requirements should be obtained. To produce a 3D model, the UAS collects overlapping images and calculates elevation based on these images. The distance of overlap should vary according to the level of detail in imagery collected. With a high level of variance in texture, less overlap is required and a higher flight elevation is preferred. Areas of low variance in texture such as snowy, sandy, or uniform vegetation then require a higher overlap in images collected. Since the Litchfield mine is mostly uniform, the UAS was flown with a high image overlap to extract common characteristics between images. For the area, a frontal overlap of 85% and a side overlap of 70% is adequate. See Figure 1 for the flight path, imagery captured, and photographs processed.

Figure 1: Pix4D flight path represented by green line, images collected as red dots, and red line as images processed

A rapid check can be performed to determine whether sufficient coverage was obtained by the image overlap. This can minimize risk at the expense of a fracture of processing time.

Pix4D can process multiple flights as long as the flight elevation is consistent among flights. Oblique images can also be processed with a high resolution of point cloud information. Although ground control points (GCPs) aren't required, they are extremely useful in any data where elevation needs to be accurate.

In Pix4D, a new project should be created including the date, location, platform used, and capture elevation to ensure all data is represented and accounted for accurately. Field information during collection is also important while capturing and is essential in processing accurate outputs.

Imagery data was batch imported using the add directories method (Figure 2) because the data was already delineated to remove non-essential photos in processing such as images before the platform reached its flight altitude, areas over water, and other factors limiting accurate returns.

Figure 2: Batch importing UAS imagery using the add directories feature

Evaluating the default camera settings is also essential. For this flight, the shutter mode needed to be switched to the linear rolling shutter option (Figure 3).

Figure 3: Editing the default camera shutter mode

After these parameters were set and the flight path was in place, processing options was then selected to make ensure proper processing. For this path, the only function switched was the raster digital surface model (DSM) to be switched to a triangulation method which has produced better results in the past.

The initial process was first ran to ensure the data would be properly processed without error (Figure 4). This can ensure accuracy without having to wait for processing of the point cloud and mesh, and the DSM, orthomosaic, and index to be completed.

Figure 4: Initial process in Pix4D displaying photograph elevation and overlay

Following completion of the first processing, the next two processes were performed. After completion, a quality report was then generated to visualize overlay and images used and preview outputs (Figures 5, 6, and 7).

Figure 5: Flight path and images collected shown in the quality report after processing

Figure 6: Image overlay along flight path shown in the quality report after processing

Figure 7: Preview of the orthomosaic and the corresponding DSM before densification in processing

After the processing was completed, an interactive point cloud overlay was produced (Figure 8). For this project, the 3D model can then be used to run quick and painless volumetric calculations to be used for asset inventory. 

Figure 8: Completed point cloud overlay in Pix4D of the Litchfield mine

RESULTS/DISCUSSION

Figure 9 shows the orthomosaic of the 197 photos captured in the UAS flight. As discussed in the methods section, a high overlap and flight elevation was used to properly display the uniform texture. This produced a high resolution overlay for further analysis.

Figure 10 then shows the DSM overlaid on the orthomosaic. This is symbolized by elevation on a stretched color scheme. The elevation produced is based on an ellipsoidal earth method that is recorded from the platform sensor as imagery EXIF information. This is different from typical elevation data which is measured referencing mean sea level.

Figure 11 shows a hillshade model of the Litchfield mine using the DSM created above. This gives a more simplified model of elevation using shadows in place of a stretched color scheme.

CONCLUSIONS

Point cloud information can produce accurate 3D models in a cost-effective manner. It is expanding into many different sectors for a wide variety of uses and will only become more popular because of its efficiency and cost-effective process. For the Litchfield mine, volumetric analysis can be accurately and efficiently made producing asset inventory data. To produce accurate information however, proper settings should be in place before collection of data such as flight elevation and image overlap. Special note of the platform, camera, and collecting methods should be taken into account during processing. Here, meta data is key to ensure accurate collection, processing, and modelling is performed.