Processing settings DJI Phantom 4 RTK

Hi Pix4D,

we’re currently in the progress of figuring out the best capture and processing settings for the DJI Phantom 4 RTK. A couple of questions:

Data capturing

  1. The P4RTK offers the choice between saving undistorted or corrected images. Which mode do you recommend for Pix4D?

Once the data is captured and the images are correctly geotagged using either RTK or PPK, what are your recommended settings for processing the data in Pix4D?

Image Properties Editor

  1. DJI stores camera model settings in the image meta data. Do you recommend using those for a newly defined camera model or use the pre-defined and detected camera settings in Pix4D?
  2. Which vertical and horizontal accuracy should be defined?
  3. Should the readings in Omega/Phi/Kappe be altered? 

Initial Processing

What settings should be used in Step 1 - Calibration, especially regarding

  1. Calibration method
  2. Internal Parameters Optimization
  3. External Parameters Optimization
2 Likes

I am also looking for Image properties setting recommendations.

With RTK images, we change nothing at all for processing which comes great reuslt.

But with PPK images, I would like to know what settings should I used for horizontal/vertical accuracy as well as the Omega/PhI/Kappa settings.

Pix4D staff, any resolution to Philipp’s questions? I am also utilizing the P4 RTK and am experiencing a myriad of issues. If  some of the aforementioned issues can be resolved I can create a second thread covering the others. I would hate to duplicate requests for information.

Hi All,

More on the optimal workflow when processing DJI Phantom 4 RTK datasets can be found in this community post:
[Desktop] Processing DJI Phantom 4 RTK datasets with Pix4D

Best,

@Phillip, we recommend using the original (undistorted) images when processing with Pix4D.

The optimized camera parameters in the internal Pix4D camera database are calculated for the undistorted images and you might face processing issues when using the corrected images.

@Blaz (Pix4D) Please provide the following: 

Data capturing

Once the data is captured and the images are correctly geotagged using either RTK or PPK, what are your recommended settings for processing the data in Pix4D?

Image Properties Editor

  1. DJI stores camera model settings in the image meta data. Do you recommend using those for a newly defined camera model or use the pre-defined and detected camera settings in Pix4D?
  2. Which vertical and horizontal accuracy should be defined?
  3. Should the readings in Omega/Phi/Kappe be altered? 

Initial Processing

What settings should be used in Step 1 - Calibration, especially regarding

  1. Calibration method
  2. Internal Parameters Optimization
  3. External Parameters Optimization

@Connery, the general recommendations can be found here: [Desktop] Processing DJI Phantom 4 RTK datasets with Pix4D.

As for the “Should the readings in Omega/Phi/Kappe be altered?”:

  • We recommend keeping the values that are automatically imported from the EXIF when creating the project in Pix4D. 

Let us know if something else remains unanswered.

Maybe I missed the answer but it’s still not clear to me what to enter in the horizontal and vertical accuracy columns in the image properties editor.  I’m using the Propeller PPK solution.  They provide RMS error values for each image in X,Y,Z coordinates.  There doesn’t appear to be a way to easily import these values using the FROM FILE function.  I’m able to enter values for Horizontal and Vertical accuracy for all images.  It doesn’t seem practical to manually edit the values for each image.

What value should I enter?  The mean RMS error?  The max RMS error?  I’m concerned that entering a value less than the max will improperly weight the image position during the processing.  At the same time I don’t want to underestimate the accuracy of the image positions because the whole point of using the P4RTK is to reduce the  number of GCPs while maintaining final accuracy.

Seems like it would be nice if Pix 4d would introduce a function to import the X, Y, and Z values for accuracy of each image using the CSV.

Here are the results I’m getting with 1, 2, 3, 4, or 5 GCPs. Given these results I will probably be using at least three GCPs going forward.  I ran another test where I positioned my 3 gcps very close together rather than distributing them around the edges of the site.  The checkpoint results were still very good.  Better than 0.1 feet.

Hi Brian,

When flying with P4RTK there are two possibilities, RTK and PPK mode. While the “RTK mode” directly adds refined accuracy values in the EXIF the “PPK mode” writes the standalone values with a few meter accuracy. 

Could you have a look at the Phantom 4 RTK - PPK processing community post and let me know if it answers your question about the PPK procedure?

We are already considering implementing a new option where it would be possible to specify what each column represents when importing the image input files. This way the input image file would not be limited to the predefined structure. 

Could you let me know if the values in the attached spreadsheet are in meters or in feet?

Best,

1 Like

Blaz

Thank for responding.  I’ve read the Phantom 4 RTK - PPK processing post several times.  The process I’m using with the Propeller PPK processing uses one Aeropoint which acts as a base station for the PPK and provides image locations within about 0.1 feet.  For each image they give me RMS error information.  What I’m looing for is either a way to enter the actual accuracy values for each individual image in the X, Y, and Z coordinate or direction from Pix 4d on what I should enter if I’m going to select one value for all of the images.  Should I enter the mean value of the RMS error for all images?  Or should I enter the maximum value for any one image?  Something else?

The values in the spreadsheet are in feet.  I only posted that in case anyone was interested in the results.  I probably should have left it out so I didn’t distract from my question.

Brian.     

Correction.  I hadn’t seen that post that you linked to.  I’ll take a look.

Here is a sample of what we are provided by Propeller.  No orientation data.

Thank you for attaching the screenshot as this way I can get a better overview of the situation. 

The orientation is written in the EXIF of the images and that is why I would first import the images into Pix4D and use the “To File…” option that will export both image geolocation and orientation. In the further steps, it is then possible to combine the .csv file generated in Pix4D with the one from Propeller Aero.

I provided the reply in the Phantom 4 RTK - PPK processing as it is more related to the question. If it is OK for you, we can continue the discussion about the workflows related to the PPK procedures there. 

@Blaz

Can you clarify this comment?

@Phillip, we recommend using the original (undistorted) images when processing with Pix4D.

Do you mean use the original (distorted) images? In other words, use the images taken with “Distortion Correction” turned off?

If I understand it correctly;
Original=Distorted
Corrected=Undistorted

Maybe I’m just interpreting the terminology incorrectly.

Dear Steve,

Pix4Dmapper (and most of the photogrammetric software packages) work better with original images that do not have any correction algorithms applied. Photogrammetry relies on the reconstruction based on common features found in images, which are then transformed into keypoints and enable the stitching. Pix4Dmapper computes the distortion correction based on thousands of tie points which are generated automatically in Step1 and it is computed for all of the images at the same time, therefore there is a lot of redundancy.

Long story short, you just need a good image dataset (good lighting conditions so that the software detects the features correctly in each image) with good overlap, a high spatial resolution (i.e. small pixel size/GSD) and with a highly accurate geopositioning (if desired). As emphasized, I would not include the geometric correction for the image dataset.

Hopefully this helps.

Cheers,
Teodora

Link is broken

HI @Mark_Turrin ,

Could you try it now? For me, it works already: Processing DJI Phantom 4 RTK datasets with Pix4D.

Thanks

Ok Is working now.

1 Like