Support Website Contact Support Blog

Weed mapping

Today we made a flight with the sequoia to scout for weed signatures on 40 acre field. We had a very positive signature of the weed plants over a very scattered grass covered soil using SIPI vegetation index. Our mais goal now is to generate a weed map to upload in a gps guided machine to make a precise aplication of pesticide. Any ideas for this workflow?

When generating a prescription map on the fields using this spot located signatures we are having awfull results.

Does anyone already have a workflow for making a precise spot pesticide aplication shapefile using a vegetation index?

Thanks

2 Likes

is it possible to make a index formula using a specific wavelenght band interval?

I.e: 800nm−505nm/ 800nm−690nm

I know this is SIPI2 formula, but I would like to obtain a different signature for weed plants, maybe shorthening this intervals a bit

@sesekeh180 800nm−505nm/ 800nm−690nm would translate to NIR-green/NIR-Red for a sequoia. For narrower wavelengths, you would need a hyperspectral camera or a multispectral camera that is custom made with these wavelengths.

image

Can you give a screenshot of the zonation map in Pix4DFields along with the SIPI index which you are using to generate the prescription map? If you are more comfortable sharing the data personally, you can also write us a personal support request (click on contact support here: https://support.pix4d.com/hc/en-us). Normally, I would use a boundary for the field and remove roads, buildings etc if they are in the map.Then I would create the zonation map. How many zones/classes are you using?

Hi Momtanu, By using this formula you wrote I am having really different results than the original SIPI2 formula presented in the software. This being said, can you provide me the SIPI2 formula used by the software?

Hi Gustavo, We use the same formula. You can see here: https://support.pix4d.com/hc/en-us/articles/360000891623-Generate-a-vegetation-index

I will generate both results and post here for you to check the differences, just a moment

thi first image was obtained by the formula you wrote me, the second image is the SIPI2 from the software, the third image is the zonation generated over the SIPI2.

when on the field we were able to identify weed signatures in a very realiable way by the use of the SIPI2. our goal is to generate a 2 ou 3 zone prescription over this SIPI 2 map. where the green areas would not need any amount of aplication (zero) and the yellow and red areas would be where we would focus the use of pesticide.

here you can see we used gorund marking points to later check if we would be able to identify in a positive way this three weed plants. and we did it, using the SIPI2. but when trying to create a zonation map to precisely aplly pesticide over this spots, we are having awful results.

I can understand. I think it would be good to use more number of zones. Do trial and error and select 4-7 zones. See which works the best. Then you can only provide pesticide on the zone where the weeds are there. Let me know if it works.

Pix4DFields uses the formula (NIR − GREEN) / (NIR − RED) for SIPI2 but clamps the values so that the variability is better visualized. What is your range in the histogram when you create SIPI2 manually?

About the number of zones, I have already tried it all the options available. It simply does not work.

the range of the manual SIPI formula was from 0.89 up to 2.4.

The most frustrating of it, is that we were able to positively map the weeds precisily using the SIPI2 Index, but we can’t generate a product to get rid of them. If we are trying to make precise agriculture, we must have precise results, in order not to waste money using pesticides in areas that it is not needed.

Hi Gustavo, I can understand. I will give your feedback to the developers and maybe we will be able to suggest something, let’s see. The reason you are getting SIPI range different in your formula is because Pix4Dfields is not clamping the values from 0-1 when you use your custom index. In the formula generated by Pix4DFields, we use the same formula but clamp the values between 0-1 for more variability.

Gustavo, Did you try with quality level “high” in the zonation map options?

Yes I tried. It does not work.
Also, we should be able to clamp this values and customize more the zonation maps.

All pix4d softwares are really expensive, you should really give more atention to the users that are trying to use your software in a commercial way.

When we ask for a feature, It is because you are not delivering what we need… if you keep turning your heads away, other software companies will come for our money…

Hi Gustavo, We have taken your suggesstion into account and passed it on to the product team. We are considering this feature for our future releases. Version 1.8 will be released sometime this month and we will not be able to integarte the feature in that, but we will ceratinly work on it for the future versions. I will update here when the feature has been released.

I would also like to schedule a call you so that we can talk more and understand your challenges better. I will reach out to you directly through email.

Gustavo, I have sent you a mail from Pix4D Support. You can directly reply me there. If you did not get the mail, you can suggest anytime next week for a quick meeting. Thank you.