How to use DJI Phantom 4 Multispectral (in-field video tutorial)

Community Input:

  1. Have you tested this workflow?
  2. What works?
  3. Other suggestions?
2 Likes

Hi Nathan,

Thank you for this great video that explains every steps involved in the process of generating orthomosaic maps.

I was wondering how critical the calibration images are. I do not own a calibration panel as it did not come with the drone. I never calibrated my images in 2021 and I was wondering if I should try to buy a calibration panel to ensure consistent maps for my 2022 season. I had in mind that since the drone has a Integrated Spectral Sunlight Sensor, the calibration was not necessary, am I wrong with that thought?

Cheers,
William

Additionally,

Do we need to change our course angle every time we go into the field? I saw that when I adjust the course angle to what is optimal, the orthomosaic map comes out with weird pattern as the course was oblique to the field, whereas when I keep the course angle parallel to the field, I don’t get these weird pattern in my orthomosaic map.

Thank you for your help!

After a fait bit of reading, I would like to use the MicaSense Calibration Reflectance Panel to perform calibration. Do you know if we can use this panel with a Phantom 4 multispectral directly within Pix4dFields? From my readings, it seems that the MicaSense Panel can only work with the Parrot Sequoia and would need additional information to work with a Phantom 4 Multispectral as the red, red-edge and NIR wavelengths are not the same between the Sequoia and the P4M. Is that still the case and if so, how can I find these values without having a spectro-radiometer with me?

Thanks for your help!

Hi William,

Great questions.

The calibration target is not necessary, however it does provide a stable reference/standard by which our software can calibrate the data on that particular day in your specific area. The integrated sunlight sensor provides important data about the sunlight. However, without very detailed internal calibrations it is very difficult for most cameras to be 100% targetless workflow. This is why many leading sensors include one.

On your second question, course angle changing is something DJI has implemented. It is said to improve the performance of the light sensor itself, providing a better view of the sun above. If you are getting a strange pattern that may be an impact related to pixel size and row width. I would suggest using the new blending option in version 1.11 which may solve that problem. Another tool now is Accurate Processing in 1.11, often times it can correct issues with DJI Phantom 4 Multispectrals when flights are performed at a high speed.

Thirdly, if you want to use a MicaSense calibration panel for the Phantom 4 Multispectral you can. It is fully characterized for each band and works well. The Sequoia panel only works for a Sequoia as it was only characterized for it. Another target which works - the one which I use in the video - is from Sentera and it also works well. We do not sell them and you would need to contact these vendors or dealers separately.

Thanks for your questions.

  • Nathan

Hi Nathan,

Thank you for your quick response!

So if I understand correctly, if I want to calibrate my Phantom 4 Multispectral camera, the Sequoia panel wouldn’t work with the multispectral camera included on my P4M. As for the Sentera calibration panel, it seems impossible to find one online unfortunately. I will continue my search, hoping to find a retailer that has one!

Cheers,
William

UPDATE TO VIDEO: DJI has recommended to use a specific speed of microSD to improve the performance. Specifically: use the SanDisk Extreme Pro A2 V30 microSDXC UHS-I for internal daily flight. However in the end users can use any microSD card with V30 speed standard.

2 Likes

Buenas tardes Nathan.
Me podrías indicar cuáles son los parámetros de calibración de la cámara del P4M para aplicarlos en el pix4dfield?
Muchas gracias.

Hello David, PIX4Dfields will automatically perform the calibration process based on the available information on the EXIF of the raw images and also with reflectance panel images if it is provided. You don’t need to insert the calibration parameters in PIX4Dfields. PIX4Dfields will automatically detect the camera using the metadata of the raw image and use those values. For more information, visit

Hi Nathan,
Are you setting the reflectence value in Pix4D for the calibration plate. We used to run the Micasense RedEdge M, and we would select the calibration images in Pix4D, draw a box on the calibration square, and set the reflectance value.

Can this be done with the Senetra calibration plate, and if so, what are the calibration values for each band?

Hi Daniel,

The target acquisition is automatic in Fields, so there is no need to manually do that part unless you have a sort of older target that doesn’t have QR codes for some reason. No need to write these manually.

Thanks Nathan.
The P4M has slightly different bandwidths compared to the Sentera 6X (below), so are the calibration values still correct for the P4M?

Lens Blue Green Red Red Edge NI-R
P4M 450 (±16) 560 (±16) 650 (±16) 730 (±16) 840 (±26)
Sentera X6 475 (±30) 550 (±20) 670 (±30) 715 (±10) 840 (±20)

Also, we’d be using Pix4DMapper, not fields, so if you have the correction values, they would be greatly appreciated.

Thank you

Hi Daniel,

Good point, the two different sensors indeed don’t have the same exact bandwidths. However the response on the target is fairly flat. We have the function programmed into our autodetection scripts. It should be working with Mapper as with Fields. Have you tried in Mapper? Have you tested with a trial of Pix4Dfields?

Is it possible to use the phantom 4 pro 2.0 for anything agriculture related?

Hi Cameron,

Yes indeed there are many uses for the Phantom 4 pro 2. You could detect weeds, terrain analysis (drainage/irrigation), field boundary creation/validation, scouting, etc. etc.

Have you tried mapping with PIX4Dcapture and processing in PIX4Dfields?

I haven’t used PIX4Dcapture but I did use drone deploy and then processed in PIX4Dfields (trial) and just wasn’t sure what could be made from that data.

You should be able to generate a few indices from the data, zonation and annotations are available. Export to share to cloud for scouting and if you couple that with the new satellite data (Fields 2.0) you can get NDVI and NDRE indices as well that support your RGB imagery.

Hi Nathan, thanks, We’re processing in Pix4DMapper, and getting good results. I was wondering if you could elaborate on why you recommend such a high overlap and sidelap, and a 3 second interval between images?
80% sidelap and 70% overlap seems very high considering we’re not making 3D models. Is this high overlap needed for the processing?

Similarly, does 2s interval work for images?

I understand that your settings might be optional, I’m just trying to get a feel for how much wiggle room might be in the numbers. We’re trying to cover very large areas, so any time savings will really help.

thank you

Hi Daniel,

I also produced great results with 2s shutter and 75% front and 60% sidelap, with “Capture at euqual Dist. Interval” and 4.0 M/s. We also improved the quality for the P4M in our new release 2.0.

The settings Nathan recommends are the fail safe settings, you can go way below that, but its kind of trial and error.