Calculating vegetation indices & plant/soil separation with QGIS (Drones in agriculture series, 3/7)

Подписаться 1,8 тыс.

Просмотров 15 тыс.

50% 1

In this video, we run through the steps of how to calculate vegetation indices based on RGB imagery, and then how to use this to create a classification layer that separates plants from soil and is useful for segmentation.
The equation for Excess Green is:
2 * (Green / (Red + Green + Blue)) - (Red / (Red + Green + Blue)) - (Blue / (Red + Green + Blue))
There are other vegetation indices that may be useful, such as VARI:
(Green - Red) / (Green + Red - Blue)
If you found these methods useful, please consider citing our recent paper in the journal Remote Sensing:
Parker, T. A., Palkovic, A., & Gepts, P. (2020). Determining the Genetic Control of Common Bean Early-Growth Rate Using Unmanned Aerial Vehicles. Remote Sensing, 12(11), 1748.

Опубликовано:

19 фев 2021

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 49

@edujacque 2 года назад

Thanks for the tips. I work with the same tools as you, but sometimes, we make things more complacted than they really are. Thnaks for your time publishing this.

@travisparkerplantscience 2 года назад

No problem! I'm glad you found it useful.

@JayeshUllatil 2 года назад

Travis, Thanks for the excellent Video.

@hmustafa851 2 года назад

Thank you Travis your tutorials are epic and helped me so much in analyzing the data

@travisparkerplantscience 2 года назад

Thanks Hisham! I'm glad they are useful for you! If you publish, consider giving us a citation here! Parker, T. A., Palkovic, A., & Gepts, P. (2020). Determining the Genetic Control of Common Bean Early-Growth Rate Using Unmanned Aerial Vehicles. Remote Sensing, 12(11), 1748.

@seanhill8943 3 года назад

Travis! You sir, are the man! Really appreciate you taking the time to make these tutorials. This is a game changer for me. I have been trying to teach myself QGIS for a while now to try and use my drone maps in it for data extraction but I was battling to understand a number of aspects as I could only really find stuff related to sat imagery. Your videos filled in all the gaps. Great work and brilliant paper, thank you. Tech is brilliant!

@travisparkerplantscience 3 года назад

Thanks!!! I'm glad this has been useful for you!

@seanhill8943 3 года назад

@@travisparkerplantscience Very much so! I find drones one of the most powerful tools in my toolbox and have seen the potential for their use in my profession but until now have not yet been able to fully harness the full extent of drone data in GIS software. Now that the pennies have dropped, the sky is the limit. Brilliant work and thank you!

@travisparkerplantscience 3 года назад

@@seanhill8943 Awesome! You're welcome!

@thomasperrott3064 Год назад

Hi Travis would you ever consider giving a 1 2 1 tutorial, i am really struggling over here in the UK, your videos are extremely good

@mercuriusmedia3042 2 года назад

Real helpful

@travisparkerplantscience 2 года назад

Thanks!

@longlasting6650 2 года назад

Interesting if u make video about accuration technique for calculating RGBVI also. Thanks

@leonardoscarpa8784 3 года назад

Excellent video!!! How do you do to get the RGB bands after the flights? Do you decompose the orthomosaic or do you fly in raw format? Thanks

@travisparkerplantscience 3 года назад

Hi Leonardo! If you have access to something like Pix4d, I would recommend pulling them out of the Index section of the outputs. You can see how to do that at 8:00 here: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-n3af0I_QYxA.html If you don't have access to Pix4D, there are some services that allow you to generate some RGB-based vegetation indices, such as Drone Deploy (using the VARI algorithm, info here: support.dronedeploy.com/docs/plant-health-2 )

@felicienzida7436 3 года назад

Fantastic

@travisparkerplantscience 3 года назад

Thanks!

@mehmetyuksel2497 2 года назад

Travis! these are awesome videos. Thank you so much really! I have a question for you. If you could help me, I would be so grateful. How can I calculate NDVI with RGB bands? I have red, green and blue band. But I do not have NIR. Can I assume that is Excess Green formula equal with NDVI?

@travisparkerplantscience 2 года назад

Hi Mehmet, great question! NDVI can't be calculated from RGB alone, because the equation is based on the ratios between red and near infrared light. Since RGB doesn't capture the near infrared, we can't calculate NDVI with RGB. That being said... other RGB-based vegetation indices like Excess Green are useful for many of the same applications, so you can often use them instead if you don't have a multispectral camera. The ground sampling distance is often better anyway in the RGB cameras relative to RGB, if that matters for your project. I hope that helps!

@mehmetyuksel2497 2 года назад

@@travisparkerplantscience Hi Travis, thank you so much your helping. May I ask you one more question? I have a drone image and I wanted to create a Excess Green image or VARI image. However, when I entered to raster calculator in ARCMAP, I can see only DJI photo band. I mean, I was not able to see red band and green band to calculate this equation. How can I divide and create red, green and blue band from a single RGB photo? Thank you so much really!

@travisparkerplantscience 2 года назад

@@mehmetyuksel2497 I don't use ArcMap, but in QGIS if you load an RGB orthomosaic, then go into the raster calculator, it should give you separate bands for these, such as RGB_orthomosaic@1, RGB_orthomosaic@2, RGB_orthomosaic@3, and in these, @1 is red, @2 is green, @3 is blue. You should be able to make calculations with those. That being said, if possible, it might be best to use the index values from the outputs of Pix4D or similar software. In any case, either should provide nice separation between plants and soil, for example

@claireshield757 2 года назад

Hi Travis, Thank you for the video. I have a question: can VARI be used as an indicator of plant health, or only to classify? Thank you!

@travisparkerplantscience Год назад

Sorry, my daughter was born the month you asked this and I missed it! If it isn't too late... yes! VARI and Excess Green are both related to greenness, and are strongly correlated to plant health!

@gabrielasalazar3847 3 года назад

Hi! Excellent video. Can I apply the same methodology for a multispectral image? or with the NDVI will be enough to delineate healthy vegetation. Thank you.

@travisparkerplantscience 3 года назад

Hi Gabriela, Yes! You can definitely use this methodology for multispectral/NDVI imagery. I actually use it on NDVI orthomosaics almost exclusively. An NDVI threshold of 0.5 or so usually works well, sometimes a little higher or lower based on weeds or canopy health for example. Hope it goes well!

@gabrielasalazar3847 3 года назад

@@travisparkerplantscience I will try it. Thank you so much, Travis. I am watching all your videos on this matter, and all are helpful.

@travisparkerplantscience 3 года назад

Gabriela Salazar thanks Gabriela! I’m glad they are useful for you!

@mahimayadav4537 4 месяца назад

hi sir, i have one doubt , can i calculate excess green from Landsat 8 image because i don't have drone imagery.

@travisparkerplantscience 2 месяца назад

Hi, I would recommend just using the Landsat NDVI imagery! It will give you the same basic results, if anything it will be better!

@yepindicus3015 11 месяцев назад

Hey, great video, do you have a source for the crop index you used for canopy cover? Thanks!

@travisparkerplantscience 11 месяцев назад

Hi, Here is a source for the Excess Green Index, if you are interested: Woebbecke, D.M.; Meyer, G.E.; Von Bargen, K.; Mortensen, D.A. Color indices for weed identification under various soil, residue, and lighting conditions. Trans. ASAE 1995, 38, 259-269. I will also shamelessly advocate for our paper where we published this whole workflow, including the use of excess green! Parker, T. A., Palkovic, A., & Gepts, P. (2020). Determining the genetic control of common bean early-growth rate using unmanned aerial vehicles. Remote Sensing, 12(11), 1748.

@eliodiaz7944 11 месяцев назад

Hi Travis, awesome video; we did a flight with the M3M RTK and the output was a shrunk mosaic, it had all the elements on the field but they were smaller compared to satellite imagery, do you have any idea why can this happen? kind regards!

@travisparkerplantscience 11 месяцев назад

Were the images in the correct location? About what was the ratio or factor of shrinking? It could be an issue with reading in the geotags in the wrong coordinate system if it is a major deviation

@JDawg-FR 3 года назад

Is it possible that you could share the images you use in this video or similar images to practice with? I do not have any drone images of my own yet but need to learn how to calculate ndvi from drone images.

@travisparkerplantscience 3 года назад

Hi, thanks for asking. I'm thinking about the best way to share this, and we have a website in development that should be ready fairly soon. For now, Pix4D has example data here: support.pix4d.com/hc/en-us/articles/360040354192-PIX4Dsurvey-example-dataset It may require a first step of processing the images. That processing can be done with a free trial version of Pix4D (or of course the paid version). Does that help? I can also drop a link when we get a sharing solution for these files themselves.

@JDawg-FR 3 года назад

@@travisparkerplantscience Thanks for the reply and I'll see what I can do with the link and the processing. However, when you get the website up and running I would appreciate the link to that as well. Thanks again!

@travisparkerplantscience 3 года назад

@@JDawg-FR Sounds good, can do!

@yazaaqa 3 года назад

nice video , but I have an inquiry . How can I get indexes since I have images from one drone flight but I have no indexes ,how can I get them?

@travisparkerplantscience 3 года назад

Hi, you will probably want to use software to make field-scale orthomosaics. There are several options, and the one I use most often is called Pix4D Desktop. You can see how to use it here: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-n3af0I_QYxA.html There are free trials for Pix4D and other programs that can do it for free. For Pix4D Desktop, you can pull the imagery out of the Index folders and calculate the vegetation indices using this video from there.

@yazaaqa 3 года назад

Travis Parker thanks brother i will try 🌷thanks for your quick response

@travisparkerplantscience 3 года назад

@@yazaaqa No prob!

@wirawan91 2 года назад

Please give me information. Why threshold was inputed 0.1 ? Why was not 0.20 or etc ?Thanks I've reference that threshold for ExG 0.2. Below : Ashapure, A., Jung, J., Chang, A., Oh, S., Maeda, M., & Landivar, J. (2019). A comparative study of RGB and multispectral sensor-based cotton canopy cover modelling using multi-temporal UAS data. Remote Sensing, 11(23), 2757.

@travisparkerplantscience 2 года назад

I would look at your true color images and masking layers, and compare what threshold value works best for you. Images taken in different locations may have different lighting conditions, light angle and shadows, soil color, plant canopy color, etc. and these kinds of factors may influence what the best threshold is. Once you have a value that works well for you, I would try to apply it across your entire experiment if possible.

@wirawan91 2 года назад

Well noted, Travis. Thanks for your explaining. Maybe, you can make video later about the level of accuration of this method : comparing the result of your analysis with verification on the field. How percent is the level of accuration. Thanks for on this very good information. Success for you ,,,,

@travisparkerplantscience 2 года назад

@@wirawan91 Check out figure 2a of this paper, where we compare the NDVI based on threshold of 0.5 with ExG with a threshold 0.1: www.mdpi.com/2072-4292/12/11/1748/htm If you publish with these methods, consider giving that paper a citation since it is where the content from this channel was formally published! Thanks again for all of your excellent feedback!

@wirawan91 2 года назад

@@travisparkerplantscience Well,, thanks a lot on your recommendation with your paper. Now, i'm trying to calculate canopy cover. Because i've problem to make different between cover crop and our plants. I try to use RGBVI. But, I just can make different between soil and vegetation (cover crop and plants). Even though, I want to calculate canopy cover of plant without cover crop and soil. Thanks once again on your video that is very helpfull ...

@kasunrathnayaka6 3 года назад

How to get NDVI values?

@travisparkerplantscience 3 года назад

First you will need a multispectral camera, then process the imagery with something like Pix4Dmapper or Open Drone Map, which can give you the outputs. Processing with Pix4D is explained here: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-n3af0I_QYxA.html I don't have a video explaining Open Drone Map but I am hoping to make one. If you don't have a multispectral camera, you can use RGB-based vegetation indices to do a lot of the same things

@kasunrathnayaka6 3 года назад

Thank you....