Introduction to PIX 4D, with GCPs

 Introduction

Pix4D is an application for processing drone imagery and creating various types of orthophotos. The orthophotos can be used to analyze the imagery in a 2d map or a 3d imagery. It can be used with all sorts of sensors and methods of capturing imagery, from a simple DJI Mavic pro to a Sony a6000 with a PPK GPS for positing. In this lab, a few sample datasets were processed to aid in the introduction of this application. 


Methods

In this lab, data captured on the Purdue Wild Life Area was processed. This data was taken with various aircraft and was all processed in PIX4D. This lab served as an introduction on how to process the different aircraft data as they each have a different set of parameters. 

Mavic 2 Pro

The Mavic 2 Pro was flown in a grid pattern, and it captured 81 images that were stitched into an orthophoto in Pix4D. The Mavic Pro 2 has a camera with a rolling shutter. Hence, it is important to verify that the automatic camera settings have applied the linear rolling shutter to the shutter model. The data would output wrong if it were set to a global shutter. The coordinate system was verified to be WGS 84/ UTM Zone 16N, we started processing a 3D map. The 3D map processing took about 3 hours as it takes Pix4D longer to process a larger quantity and quality photos. Figure 1 shows the flight path that the Mavic Pro 2 took, with the red dots being the points that a photo was captured at. This can be used to ensure that the flight path looks acceptable before running the initial processing phase. Next, the initial processing phase was run, and the quality report was inspected. Everything in the quality report was acceptable, so the point could and the mesh was run next. Figure 2 depicts the imagery's ray cloud view, which shows all the camera angles at the top of the imagery. The pixels on the lower part of figure 2 are tie points for the imagery, and that is where parts of various images overlap and are stitched together. Processing in the DSM (Digital Surface Model), orthomosic, and Index are processed, and figure 3 depicts the results of the processing. This is a 3d map of the area, and it could be analyzed for a varsity of applications. 
Figure 1. Flightpath of the Mavic Pro 2

Figure 2. Ray cloud view of the Mavic Pro 2's Imagery



Figure 3. Point Cloud and Triangle mesh view of the Mavic Pro 2's Imagery

                    
Figure 4. Zoomed-in view of the Mavic Pro 2's Map


DJI XT2

Next, data were processed from the DJI XT2 camera, which captures both RGB and thermal images. The two types of data output from the XT2's imagery makes us run 2 processes to process all the imagery. For the RGB imagery, the same process described in the Mavic 2 section is completed, and an orthomosic is created. Figure 5 shows the orthomosic created by this mission, and figure 6 showed a zoomed-in view of the map. There is a quality difference between the Mavic 2's orthomosic and the XT2's orthomosic. As seen in figure 4, Mavic 2 has more detail than the XT2. This is due to 2 reasons: the Mavic 2 has a better camera, which can capture images at 20 megapixels, whereas the XT2 can only capture RGB images at 12 megapixels. The megapixel difference could not be if they have the same pixel per cm resolution, but there is another factor, the number of images processed.  The Mavic 2 captured 81 images, whereas the XT2 captured 15. The increased number of images captured by the Mavic 2 allows for more resolution from different angles. 

Figure 5. DJI XT2 Orthomosic

Figure 6. DJI XT2 Orthomosic Zoomed in view

The thermal data was also processed into an Orthosmosic for the thermal imagery collected by the XT2. The only difference in the setup was a change of template from 3D Maps to Thermal Camera, which generates a reflectance map. This was done as the way that Pix4D processes these types of imagery are different. The orthomosic cannot be seen in Figure 7 because it is created a .tiff file, but this can be imported into other applications such as ArcPro GIS to be analyzed. 

Figure 7. Thermal Point cloud XT2

Sony A6000 

The final orthomosic that was generated was with a Sony A6000 camera and a PPX. The Sony A6000 does not record GPS information in the image's metadata, so we had to import PPX data from the aircraft that correspond to the images. The flight map in figure 8 had me worried about the PPX points, as it is not a flight grid that I would expect to see when processing orthomosics. In this case, I would contact the pilot to verify the flight plan as it is a nonstandard route.  But for this lab, I proceeded and processed the orthomosic, and it came out acceptable. Figure 9 shows the raycloud from the A600 flight can be seen, and it looks similar to the other flights prior. 
Figure 8. Sony A6000 flight Map



Figure 9. Sony A6000 Ray cloud

Ground Control Points

This mission included 6 Propeller arrow points, a commercially available ground control point (GCP). The GCPs were used to correct the orthomosics to higher accuracy than without them. This was completed for each of the sensors RGB data, the thermal camera could not be used with the aeropoints, as they have the same thermal temperature across the GCP, and an exact center is hard to find. 

For the datasets GCP was not included as it had high variance, and was located under a tree, so even if one wanted to include it in the data set, it would be hard to click on as it was hidden. 
Figure 10. GCP Variene for GCPs 1-6
Figure 11. GCP 6 is located on the left side of the image near the gravel road, and can be seen to be under a tree

All three of the sensors used in this lab have different resolutions, in figures 14 and 16, the difference in the capability of zooming into the GCP for selecting the center can be seen, and it can increase variance for the cameras with a lower resolution. 

DJI Mavic Pro 2

Figure 12. DJI Mavic 2 pro GCP accuracy



Sony A6000

Figure 13. Sony A6000 GCP accuracy 

Figure 14. Sony A6000 GCP selection



DJI XT2 RGB

Figure 15. DJI XT2 GCP Accuracy

Figure 16. DJI XT2 GCP selection



Quality Reports

The Quality reports were all analyzed and looked to be fine, so all of the GCP adjustments were done properly. The main sections that were checked over were the quality check, the calibration details, and the geolocation details. In the Quality check, this is where the first sign of an issue would be, I ran into a few problems while running the A6000 GCP corrections, but after adjusting the center of the GCPs to a more precise location, I got all green checks.

Conclusion

Pix4D has been a useful tool for creating orthomosics with drone imagery and processing that data. I personally have been using it occasionally since 2014. I have found it a beneficial tool to use on certain UAS projects, even though the processing aspect can be a time consuming one. The GCP correction to the data improves the quality of the data processing and orients the orthophoto to the basemap closer, as well as can be used to overlay multiple orthophotos with out having the maps be shifted. 

Comments

Post a Comment

Popular posts from this blog

Setting up a vtol in mission planner

Image
Introduction Setting up a vertical take off or landing (VTOL) uav is one of the more complex aspects of mission planner, for programming a pix hawk flight controller. The vtol that will be set up in this blog is a 2.5 meter wingspan, custom quad rotor, with all four motors tilting uav. Figure 1. A screen capture of the firmware installation page of mission planner  To begin install arduplane on your pixhawk, using the install firmware page in mission planner, figure 1 shows the pop up window that should appear. Next once the firmware is installed go through each of the settings of the mandatory setup, including the accelerometer calibration, compass calibration and the radio calibration, but skip the servo output for now. The next step is going to the full parameter list under the configuring and tuning tab, and enable Qplane, which stands for quadplane, and allows for all the vtol settings to be shown. Then the frame style and setup has to be configured, these have t...
Read more

DoakBurn (Again)

Image
 Introduction The goal of this lab was to use ArcGIS Pro to view the various band stacks of orthomosaic imagery and a normalized difference vegetation index (NIVD). The data was collected by Dr.Hupy, William Weldon and Zack Miller, and the data was of a controlled burn with multispectral cameras. Methods The methods of this lab was to look at the areas that were burned and compare them to the areas around them that were not burned. This could be done with RGB, or band stacking for the imagery, which allows the ability to see things that are not able to be seen with the human eye.  Metadata from the Flight Vehicle: Bramor PPX Sensor: Altum Flight Number: 2 Takeoff Time: 12:18 Pm Landing Time: 12:35 Pm Altitude (m): 121 Sensor Angle: NADIR Results  The data collected with the band stacking shows the differences between the pre-burn and post-burn data. Figure 1,2,4,5,6,9 are in a 5,3,2 band stack, meaning the red represents near-infrared, the green represents th...
Read more

Designing a Vtol Mechanism from scratch

Image
Introduction A project that I have been working on needed a Vertical takeoff or landing (VTOL)  tilt mechanism for the the front 2 motors, so here is how I went about designing and testing it. It can be challenging to hand launch a twin motor plane as you have to worry about both prop disc areas, while managing controls, and adding a vtol system can allow for operations where a wheel based takeoff is not a possibility  The specifications of the propulsion system are a Emax 2213-935KV brushless motor, that spins a 8045 prop, with a spider 20apm opto esc, using a 4 lithium polymer cell battery. This power systems is typically used in quad copters such as a dji flamewheel 450. The servos for tilting the motor varied depending on the design of the tilting mechanism. Figure 1. A dji flamewheel 450 quadcopter, with a propulsion system similar to the VTOL's system Goals  The goals of this project was to be able to develop a vtol tilt system for a uav that ...
Read more