Bands in NDVI image matching flight path

My pix4d output almost always show bands according to the flight plan. I see it as well by just doing normal processing (non-ndvi).

 

Why would this be? The validity of my NDVI results are in question because of this. I’m not flying when cloudy, etc. Do I need more overlap? What can cause it?

 

I’ve seen others experience this problem before but I can’t find any solution.

 

Can anyone give pointers?

The validity of your NDVI results SHOULD be in question because of this. 

 

What camera are you using? Many manufacturers have software to preprocess images before running them through pix4d.

 

Environmental/acquisition problems can be varying light conditions and (very) commonly the pitch & roll angles of the aircraft. 

Sensor problems can be vignetting, non-locked exposure & WB settings, etc. 

I’m using Tetracam’s ADC Snap. There were no clouds, I have a roll gimbal keeping the camera perfectly horizontal - pitch is negligible.

 

I’m thinking maybe it’s glare from the Sun - creating bright reflective spots while overhead?

What was the flight duration? 

The sun spots usually don’t create the consistent striping effect. 

 

Did you follow the guide here for tetracam projects? 

https://support.pix4d.com/hc/en-us/articles/202560219-How-to-create-and-process-a-Tetracam-dataset#gsc.tab=0

 

 

Flight was about 15 minutes max, perhaps 70 photos.

 

Yes I followed the guide it exactly.

Here is another flight, this time I’ve used Microsoft ICE to compare and it’s clear that there’s a problem with the Pix4D render:

Uh, no.  Your goal is to create something visually appealing. What your goal SHOULD be is to create something accurate. 

Hi Derrick - I don’t give a hoot about aesthetics! From the images it’s clear that the data is not accurate at all. The left image shows a rendering problem. I can’t give this to the client as it’s clearly incorrect. Those bands follow my flight plan precisely, across many lanes of crops - so it’s definitely not reality.

I agree - it shows that the data is not accurate at all, and you cannot give this to a customer. 

MicrosoftICE is not the correct way to find the problem, and it does not prove that Pix4D is the culprit. 

 

Have you tried comparing reflectance values in each mosaic to those of the raw images themselves, before processing? 

I have, but individually the images’ albedo is very subjective and Tetracam does not seem to see a problem. The camera is set (as recommended) to auto exposure - which is a red flag for me - yet that seems to be the “best practice”.

What I try to illustrate on MS Ice is the difference in brightness of the source images are not as pronounced as it appears in Pix4D. I’m thinking maybe there’s a setting in Pix4d that calibrates the images, but something goes wrong there.

 

Either way, I’ll do another flight, this time with auto exposure off.

 

Thanks for your input.