Hi,
i came across 2 severe Problems with processing my Multispectral Images shot on a Phantom 4 Multispectral with Pix4DFields. (at the Moment i am still on Trial)
First off the Camera and Workflow Settings: (were used over all Photos shown)
Overlap Front and Side: 80%
Photos were taken at equal Time Intervals.
Capture Time was set to 2sec
AGL was 100m.
Weather Conditions were cloudy.
RTK Positioning was performed with the Base Station from DJI.
Rig-Calibrationpoints in Pix4DFields were enabled.
IMU, Gimbal, Compass and Firmware were calibrated Pre-Flight.
Second the disred use of the photos:
The Photos will be used in professional Agriculture. Therefore a AGL of 100m, the equal Time Interval Trigger and a Time Interval of 2 sec was selected to achieve a better Flight Performance per Battery.-Yes i know this comes with less Precision but everything apart from these Flight Options is simply unacceptable to gain a sustainable Rate of ha/hour.
The 2 Problems i am facing right now:
1. - When creating the Orthomosaic i came across very blurry spots in the Photos that were following the lines of Flight created by me. They seem to follow the outer Parts of the Pictures taken.
2. - The Individual Bands didnt overlap enough in the Multispectral Image, making the Pictures seem blurry and shifted. Its becoming very obvious when i create an RGB from the Single Bands.
Similar Results were achieved at the following Heights (50,25), the following Overlaps (90/90, 70/70), and both other Flight Options (Hover and Capture(even at 3,6km/h), Capturing at equal Distance Intervals 1-2sec).
Am I just expecting too much from the system? Why dont the Single Bands overlap precisely enough? What can I do to improve?
I am attaching Photos to compare. When comparing the NIR and RED-EDGE there is visual Shifting/Morphing.
Any Help is welcome.
Thank you very much in advance!!
More Examples:
3 Pictures taken closely together. 80/80% Overlap. All 5 Bands included.
Red Edge Orthomosaic is completely fine.
RGB-Picture blurry
1 Like
Hi,
Thanks for sending this very detailed issue, we have forwarded it to our team and we would like to see the dataset to give you a better answer. Could you please upload it here? Phantom 4M Dataset
Could be from several heights, overlap as you prefer.
Best,
José
Hi,
Thank you for the quick Reply!
I am uploading the Photos now. The Folders are labeled as to what the Missions Parameters were.
Best Regards
Hi,
Thanks for sending the datasets.
One one hand, I would recommend flying with three rows over the field, I see there are mainly just two. You can see this video about the flight planning PIX4Dfields - How to use DJI Phantom 4 Multispectral (in-field tutorial) - YouTube
On the other hand, this is an issue that we hopefully will address on next PIX4Dfields release. It is related to a async. rig on the camera.
For the time being, I have processed your “Agl100m_EqualDistanceInterval_Overlap9090_Shutter1seconds” dataset successfully on PIX4Dmapper(you can have a free trial, https://www.pix4d.com/product/pix4dmapper-photogrammetry-software). You can find the results on same drive.
To do this:
-
Just select multispectral images, Select Multispectral Template, after go to: Project > Image Properties Editor…, and select “Optimize Independently”. Please read Processing camera rigs
Please submit us a ticket at support if problem persist so we can track it better.
Regards.
Hi,
Thank you for the processing.
Right now i cannot create orthomosaics in the trial version.
I guess (since you have the Tutorial Video up) the company has at least some experience with processing pictures from a P4M. Is this a common problem or is my rig just problematic? I do have the option to return if it is not in the desired shape.
Are there any options to try the functions without purchasing the license?
Thank you in Advance.
Hello,
Could you please open a support ticket using this link
https://support.pix4d.com/hc/en-us/requests/new?
Best,
José
Hey,
If anyone is facing the same problems:
processing the dataset in Pix4DMapper worked out just fine.
Best Regards
