Hello,
I am wondering if Pix4D mapper is compatible with the new DJI Phantom P4 multispectral?
Hi,
DJI P4 multispectral is compatible with pix4Dmapper. It was released in the version 4.5.3. You can see our technical release notes here: https://support.pix4d.com/hc/en-us/articles/115005738363-Pix4D-Desktop-preview-and-technical-release-notes
Thank you very much for your help! I will check this out and let you know if I was able to process phantom 4 multispectral images.
Cheers!
1 Like
Hey, were can i find a full workflow?
Hi @yourulf, The workflow is the same as any other multispectral image processing in Pix4DMapper. You can have a look here: https://support.pix4d.com/hc/en-us/articles/209362146-How-to-process-Sequoia-imagery#label1. If you are looking for any specific answer or having any issues with the workflow, please let us know.
i still have some problems my images. still there are artifacts. Is the irradiance sensor data of the p4ms used in pix4d?
` Would love to talk to an engeneer to understand that issue better, i was digging into the metadata of each tiff but i am not sure were the information is stored.
You are talking to one 
The tags are stored in the EXIF, you can use exiv2 to read the tags. The irradiance and sun angle tags that Pix4D uses are Xmp.Camera.Irradiance, Xmp.Camera.IrradianceYaw, Xmp.Camera.IrradiancePitch, Xmp.Camera.IrradianceRol. However, the sun angle tags which are the irradince yaw, pitch, roll are tagged as 0 in teh DJI p4p images.
It depends on you whether you want to use the light sensor data for radiometric correction. If you choose camera only correction, Pix4D will not use use the sun sensor data. If you use camera and sun irradince, Pix4D will use the DLS data.
For more info, you can have a look here: https://support.pix4d.com/hc/en-us/articles/202559509-Radiometric-corrections
Can you let us know what are these two screenshots represent? Are they NDVI (and are they both with or without radiometric correction)?
Hi Momtanu,
we are trying to process a DJI P4 Multispectral dataset (5 bands, TIFF) using Pix4Dmapper (see attached Pix4D Report). All bands are OK apart from the Red Edge. The software does not find enough Automatic Tie Points in the RE band, consequently camera calibration and orientation fail. Most (raw) RE images are very crisp and with clear details. Could you suggest a possible solution ? Should we avoid using the Sun Irradiance data?
Thank you in advance for your help,
Maurizio
P4M_28_05_20_report.pdf (1.9 MB)
Hi @Maurizio_Buonanno, Can you try with the most recent version of the software (4.5.6)? The rig was added in 4.5.3, thus you can see in your version the images are not getting processed as a rig. In that case, the software has to find keypoint from images of all bands which is difficult because red, red-edge bands have less variability than green, NIR. When processing as a rig, Pix4D calibrates all the images of only the green band and uses that to process images/reconstruct the other maps. Using 4.5.6 should solve this issue.
I am having similar issues with Phantom 4 Multi NIR and RED images. I started using P4M in May with success through June but have not been able to produce another single band image since then. I updated the software I think sometime in early July to 4.5.6. I have been using P4P for the last year with no issues and the P4M RGB appears to have no issues.
Image process status
Hi @mjackson Can you send us the logfile, quality report and p4d file?
Hi Momtanu,
I think I have similar issues with the Phantom 4 multispectral images. My NIR and Red Edge output also seems distorted. The first 3 bands, as well as the RGB- Image are processed well. The raw images are sharp and detailed enough. The Quality report says there is a 47.65% relative difference between initial and optimized internal camera parameters but the “What to do”- steps didn´t really help me.
I will attach the report and p4d file. I use the software version 4.5.6
Can you suggest a solution to this problem? Thank you in advance.
40m_att5.p4d (581.3 KB)
40m_att5_report.pdf (354.1 KB)
Hi, I see you used 0.25 as the keypoint image scale and standard calibration. I will recommend using the ag multispectral template. Let me know how that goes.
Hi,
i used 1, 0.5 and 0.25 as keypoint image scale but the results don’t seem to be accurate.
With keypoint image = 1, the results are very messed up (attached image, the top one is the red band and how it should look like. The 2 others are red edge and NIR). The results are better almost accurate as I set down the keypoint image- scale. But there are still distortions.
If I mosaic the 5 bands seperately I get good results, also for red edge and NIR. But for further work I need them to mosaic all in one run.
In the MapView there is for every Image (red points) the RGB-jpg preview when clicking on it. Except for the very first point (where my mouse is) there shows the blue band as preview (2nd attached image).
Is this normal?
Can you send me the quality report with image scale 1?
Sure, here it is. The results look better but red edge and NIR still show distortions.
40m_radcal_report.pdf (1.2 MB)
Can you process the RGB and multispectral images as separate projects? This is because the RGB camera is not included in the rig, so even if you process them with multispectral images, the RGB will not align the other reflectance maps.
Also, for the multispectral process with optimise relative rotation instead of “optimize relative rotation with a subset of secondaries” as the rig option. It will take more time but it should help with the artifacts in the other maps. With optimize relative rotation with a subset of cameras, Pix4D calibrates all the images of the reference camera (green) and only 8 random images of the secondary cameras. The primary camera images are used to reconstruct the other bands. For optimize relative rotation, all the images from all bands are used for calibration. You will find this in the image properties editor.
1 Like
Yes it´s possible to process RGB and multispectral in seperate projects.
With the option “optimise relative rotation” the results are pretty good now.
Thank you
1 Like
Hello
I have been using a multispectral phantom 4 for over a year. I use dji terra, which is interesting in some ways but sometimes struggles to put the images together correctly. I have just acquired in addition a pix4d mapper license (which I used before with the sequoia) that I want to use as the main processing software
I just did a different test on a flight, and I can’t get satisfactory results with pix4d
The drone uses a trimble rtk signal, I processed the images under dji terra and pix4d without using the gcps, I only display them to see the accuracy of the image processing with the rtk
On dji terra, geolocation is good, and image quality is good for all 5 wavelengths
On pix4d, if I use the default setting, the geolocation is good but the images are less sharp than on dji terra, especially on the red band
If I set pix4d for the multispectral process with optimize relative rotation instead of “optimize relative rotation with a subset of secondaries” as the rig option, the geolocation is no longer good (60 cm offset), the quality of the red band s 'improves, on the other hand the red edge becomes unusable and completely blurry (for the other bands this has no impact)
I have done some tests on other flights in the past, and the option optimize relative rotation instead of “optimize relative rotation with a subset of secondaries allowed to eliminate some artifacts on the images (like those shown by yourulf on the previous images), so how do you get an image as sharp on pix4d mapper as on dji terra, with such precise geolocation in cases where gcps are not used?
Hello @elebourhis, Using GCPs should improve your geolocation accuracy but should affect the sharpness of the result. It would depend on the GSD of your project and the processing option that you are using.
For more information about the rig processing option, I would suggest you to go through our support article on Processing camera rigs.
I see you have already created the support ticket, we will deal with the issue through the personal ticket and if there is anything new, I will later share it on this post.