Hi,
I have a question about my NDVI results.
For a large area that needs two flights, what is the best solution to have good NDVI values ? Is it better to do a fusion of all images ? But which images should I use for radiometric correction ? (A) or (B) in the attached image.
I tried different tests but I am really surprised by the difference between the results.
Thank you to help me.
Natacha
Hi Natacha,
For multispectral projects, it is better to process the projects separately with the different calibration target images. You can then merge them in QGIS. This is because if there is a time difference in the flights or change in weather conditions (sunlight), it is good to have different panel images.
Since only one-panel image set is supported in Pix4D, if you are procesing together as one project, the best set of panel images are recommended to be used (mostly one after the first flight and before the second flight will correspond to the weather conditions the most)
Hi, thank you but I have done two projects separately with the different calibration target images (test 2 and test 3 in the attached image up) and then, I have merged them with ArcGis.
But they seem completely different…
Hi Natacha, This looks like a big difference. However, I cannot really see the values of the index. Can you write us a personal support ticket? I would like to investigate it further. You can send me the Quality report of flight 1 and flight too along with the NDVI maps and the raw datasets of both the flights.
Hi Momtanu, thank you, I have sent you the files by a personal support ticket.
I have done more tests (see attached) and I have more questions (in red on the attached images).
I have done “Since only one-panel image set is supported in Pix4D, if you are procesing together as one project, the best set of panel images are recommended to be used (mostly one after the first flight and before the second flight will correspond to the weather conditions the most)”
Because separetely, it was completely different…
And finally, I have found NDVI=0.03 for my area with all the multispectrale images (over than 800) + the target2 images.
I still do not understand why the difference because the target images are supposed to balance the values, take into account the atmospheric conditions so the same area overflown twice at 8h30 and 9h15, should have the same NDVI. Here, we have 0.03 with the target images 2 (take before the second flight) and 0.12 for the target images 1 (take before the first flight).
But at least I find 0.03 for all combinations with target 2.
Natacha
Hi Natacha, It depends all on weather conditions. maybe the conditions changed while you flew your first flight? Maybe the condition while capturing the second set of target images was more similar to both your first flight and second flight.
Hi @Natacha, I have replied to your personal support request but wanted to update here as well. Below are 2 graphs of sun value vs time. As you can see, the second set of traget images are more similar to the weather condition of both the flights which is why we ask to capture target images before and after the flight.
There is a csv file in params folder, which users can analyse.
Thank you Momtanu for your response, I understand.
And for the image
Is it possible to do a reset and choose again the target images and do again the step 3 ?
Do the processing stay in memory because I have found strange results each time ?
If you are using the same set of target images, the result should be the same. According to the polygon you draw, there might be very small difference as the pixels in the polygon are taken into account.
After using the target images again, you will need to completely process step 3. If you calibrate and just click on index calculator, the values will stay in memory and not change.
Hi, I used the same set of target images (folder 33 that I send you). Insides, there are the target images taken before the first flight. I have seen an error of my result because the mean NDVI in my test area was 0.29 so I wanted to process again the step 3 (completely).
I used the same set of target images but I didn’t use the same images because we did 12 images but the reflectance factor value was the same for each spectral band (Ex: For RED image n° 082155_0000_RED.TIF or n°082156_0001_RED.TIF, the reflectance factor value was 0.213).
The polygon is exactly the same in the same place and size.
Hi Natacha, It does. Some targets maybe underexposed/overexposed. You should select the set of target images (from 3 captures, one capture here means 1 set of images of all bands) that looks the best to you and use that as the calibration target for radiometric correction.
When you launch the radiometric calibration with Sequoia the camera takes 3 sets of images with different exposure in order to avoid over or underexposure. Only one set of images can be included in a project. A set of images consist of one image per each band (green, red, rededge, nir). You will have to choose the best set of images (the one not over or under exposed). When the digital number of the target images are higher than 63000 we consider them overexposed.
You can read the community discussion here: Parrot Sequoia - Choose best calibration image - #12 by momtanu.chakraborty
Ok, thank you
Hi Momtanu,
I have a last question concerning index with PIX4D mapper, as you can see, we can see some effects, some light contrasts or color balance problems.
How to remove these effects that parasitize the final orthomosaics ?
If there were shadows, that can cause these issues which Pix4D will not be able to correct. This image might explain why scattered clouds can cause problems. The DLS will not be able to account for the clouds here. If you see tripes in the direction of flight lines, it might be you flew towards and against the sun. It is always recommened to fly perpendicular to the sun. We also recommend having more overlap to negate the effect.
Thank you, it was sunny, no cloud. I have tried with an other software. I will do more tests but the first result seems good visually but I have to check the values…
If you have not, you should use sun angle from DLS IMU radiometric correction.
I have tried but I couldn’t because I don’t know how to do ?
Seems like the sun angle tags were not written. These should be the tags in the EXIF. It means there is no IMU embedded in the sun sensor (Micasense has this option)
Xmp.Camera.SunSensorYaw,
Xmp.Camera.SunSensorPitch, and
Xmp.Camera.SunSensorRoll (in degree).
Pix4D uses them to do the sun angle correction. Sun angle correction usually corrects for these stripes/blobs (https://support.pix4d.com/hc/en-us/articles/202559509-Radiometric-corrections). For sequoia, Sequoia with Parrot Disco-Pro AG or senseFly eBee drones can use sun angle correction.You can also try with camera only correction and the radiometric targets. If the issue was due to shadows, then this option will give better results and you will not have the artifacts. Maybe the other software is not doing the type of radiometric correction we are doing.
You can send your full dataset to us if this option doesnt give better visual results, we can evaluate more why these artifacts are there.
It’s a sequoia camera. Which EXIF ? The EXIF of the images ? The EXIF of the target images ? Or in the .dat file that I couldn’t manage to open ?
