Alternatives to Drone Deploy

[strubler]

Are there any legitimate competitors to Drone Deploy? I'm having a difficult time reaching them to gather information. If they are that hard to get a hold of, as a small businessman, I am wondering how hard will it be to communicate when I really need some support.

 

[Keule]

Are there any legitimate competitors to Drone Deploy?

PIX4D - Professional photogrammetry and drone mapping software

 

[brett8883]

PIX4D - Professional photogrammetry and drone mapping software

Just got this thanks! I haven’t yet tried a flight with it but the interface looks promising a bit better than Drone Deploys actually. Do you know if it can do maps larger than what one battery can do? That’s a pretty cool feature of DroneDeploy that it can start where the last flight left off

 

[Keule]

Do you know if it can do maps larger than what one battery can do?

yes

Out of curiosity, do you want to use the autonomous flight apps only or do you want to use the stitching/mapping services as well?

 

[brett8883]

yes

Out of curiosity, do you want to use the autonomous flight apps only or do you want to use the stitching/mapping services as well?

For me both, I’m not the op though. I suppose I could use auto-pilot for the autonomous features and upload but it’s convenient to have the all in one package

 

[Wailuku]

From what I could determine Pix4D was the best choice for me. I teach an environmental monitoring & emerging tech course at a uni (I made up the course and this is the first semester it has run so we are very much flying by the seat of our pants). In it I'm teaching (and learning) 3D modeling, veg health indices etc etc with drone imagery. All of the academics I talked to who use drones suggested Pix4D and so far we've been happy with it.

 

[brett8883]

From what I could determine Pix4D was the best choice for me. I teach an environmental monitoring & emerging tech course at a uni (I made up the course and this is the first semester it has run so we are very much flying by the seat of our pants). In it I'm teaching (and learning) 3D modeling, veg health indices etc etc with drone imagery. All of the academics I talked to who use drones suggested Pix4D and so far we've been happy with it.

Well then I have a question for you! I have been helping another member with his mission to investigate the mass deaths of Juniper trees in Southern Utah. I suggested looking into multispectral imaging but I have noticed that drone deploy can actual convert an optical image into NDVI. Is that something that can actually be used to get any degree of insight into vegetation health?

I tried it out on some kind of pine trees that are here in northern Utah and it did seem to pretty clearly spot the dead trees from the healthy trees but will have to do more tests to see if anything else can be determined with it. What is your opinion on the efficacy of using converted optical images to determine plant heath? Is it just a pseudo-science kind of thing?

 

[Wailuku]

Well then I have a question for you! I have been helping another member with his mission to investigate the mass deaths of Juniper trees in Southern Utah. I suggested looking into multispectral imaging but I have noticed that drone deploy can actual convert an optical image into NDVI. Is that something that can actually be used to get any degree of insight into vegetation health?

I tried it out on some kind of pine trees that are here in northern Utah and it did seem to pretty clearly spot the dead trees from the healthy trees but will have to do more tests to see if anything else can be determined with it. What is your opinion on the efficacy of using converted optical images to determine plant heath? Is it just a pseudo-science kind of thing?

Just to clarify, I don't use Drone Deploy, so I can't really comment too specifically about their 'conversions' or what they might be doing differently now than when I first looked into this. And I'm not 'anti- Drone Deploy'.. I just found Pix4D to be best for my purposes.

There was a recently published white paper that looked at three of the algorithms for plant health indices based upon RGB and compared it to true NDVI (McKinnon & Hoff 2017; Agribotix white paper) and it was not a favorable comparison. That was also the general consensus I got from geographers who routinely use remote sensing data.

But, NDVI (as you may already know) is simply NDVI = (NIR - RED)/(NIR + RED). So if you don't have NIR data (which RGB cameras filter out) then you don't have a true NDVI. Thus, people develop algorithms using just the data available in RGB images to try to do as best as they can for a vegetation index with the implicit understanding that NDVI is better. So far, this hasn't been too successful (McKinnon & Hoff 2017), but that may well improve.

Two of those RGB-based indices are VARI and TGI. And while they seem to do a good job within the purposes of their original development (VARI for leaf-area index and TGI for chlorophyll(and nitrogen)), neither seem to be a good general index of vegetation health and part of the issue is that ambient light isn't controlled very well with strictly (uncorrected) RGB images from (unmodified) consumer level drones.

Now, finally to get to your actual question... if you examine VARI and TGI with an eye toward 'plant health' and compare those directly to the RGB image itself it seems to me that the RGB image is giving you the same info. That is to say, do you really even need an index if you are just looking for dead or defoliated spots? And either way, you'd still have to do 'ground truthing' to find out why those trees are dead/dying.

I hope I've answered your question.. if not I'm happy to try again or dig into this a little more if I don't have an answer myself. I'm a biologist employing some tools developed for/by other sorts of sciency folks.. so I'm still learning too.

 

[brett8883]

Just to clarify, I don't use Drone Deploy, so I can't really comment too specifically about their 'conversions' or what they might be doing differently now than when I first looked into this. And I'm not 'anti- Drone Deploy'.. I just found Pix4D to be best for my purposes.

There was a recently published white paper that looked at three of the algorithms for plant health indices based upon RGB and compared it to true NDVI (McKinnon & Hoff 2017; Agribotix white paper) and it was not a favorable comparison. That was also the general consensus I got from geographers who routinely use remote sensing data.

But, NDVI (as you may already know) is simply NDVI = (NIR - RED)/(NIR + RED). So if you don't have NIR data (which RGB cameras filter out) then you don't have a true NDVI. Thus, people develop algorithms using just the data available in RGB images to try to do as best as they can for a vegetation index with the implicit understanding that NDVI is better. So far, this hasn't been too successful (McKinnon & Hoff 2017), but that may well improve.

Two of those RGB-based indices are VARI and TGI. And while they seem to do a good job within the purposes of their original development (VARI for leaf-area index and TGI for chlorophyll(and nitrogen)), neither seem to be a good general index of vegetation health and part of the issue is that ambient light isn't controlled very well with strictly (uncorrected) RGB images from (unmodified) consumer level drones.

Now, finally to get to your actual question... if you examine VARI and TGI with an eye toward 'plant health' and compare those directly to the RGB image itself it seems to me that the RGB image is giving you the same info. That is to say, do you really even need an index if you are just looking for dead or defoliated spots? And either way, you'd still have to do 'ground truthing' to find out why those trees are dead/dying.

I hope I've answered your question.. if not I'm happy to try again or dig into this a little more if I don't have an answer myself. I'm a biologist employing some tools developed for/by other sorts of sciency folks.. so I'm still learning too.

Yea that’s perfect thanks! The reason for wanting something like this is because he wanted to have a way to track the dead trees over vast areas. Sure optical can tell you this if you have s great enough resolution but it would be labor intensive to manually count the tress in the photos. We looked into using a machine learning model to do this automatically but the model is not good at determining if a tree is living or dead. Frankly when there are hundreds of trees in a single image it’s difficult for me to determine a living teee with a dead one. However, with NDVI the contrast in the image from a dead tree to live tree is enough that the model can pick up on that difference quite easily.

 

[Wailuku]

Yea that’s perfect thanks! The reason for wanting something like this is because he wanted to have a way to track the dead trees over vast areas. Sure optical can tell you this if you have s great enough resolution but it would be labor intensive to manually count the tress in the photos. We looked into using a machine learning model to do this automatically but the model is not good at determining if a tree is living or dead. Frankly when there are hundreds of trees in a single image it’s difficult for me to determine a living teee with a dead one. However, with NDVI the contrast in the image from a dead tree to live tree is enough that the model can pick up on that difference quite easily.

I see.. in that case the TGI algorithm from RGB imagery would probably work to ID dead vs not-dead trees. But what you are doing can certainly be a labor intensive process. Many states (especially the the Pacific Northwest of the US) do annual visual surveys from planes to map/monitor tree damage (insect, virus and fungi, bears, storms etc) over the entire state (or at least the forested parts) and then conduct a bit of ground truthing to determine cause. So I guess it depends upon how vast we are talking. Also, I imagine one could build a model in ArcMap (for example) to extract by attribute based upon a TGI raster layer (or even the orthoimagery).

 

[brett8883]

I see.. in that case the TGI algorithm from RGB imagery would probably work to ID dead vs not-dead trees. But what you are doing can certainly be a labor intensive process. Many states (especially the the Pacific Northwest of the US) do annual visual surveys from planes to map/monitor tree damage (insect, virus and fungi, bears, storms etc) over the entire state (or at least the forested parts) and then conduct a bit of ground truthing to determine cause. So I guess it depends upon how vast we are talking. Also, I imagine one could build a model in ArcMap (for example) to extract by attribute based upon a TGI raster layer (or even the orthoimagery).

Thanks so much you are a wealth of information on this subject!

 

[Wailuku]

Thanks so much you are a wealth of information on this subject!

You're welcome. We could always slap a LiDAR unit on drone and do this! Two of my student groups are doing something very similar this semester, but we have really good and free LiDAR coverage here. https://www.isprs-ann-photogramm-re...net/I-7/359/2012/isprsannals-I-7-359-2012.pdf

 

[rbanzato]

Is there any tool like Drone Deploy, Pix4D etc., that is free to use?

 

[Wailuku]

Is there any tool like Drone Deploy, Pix4D etc., that is free to use?

Both of those offer free trial versions (two weeks each I believe) unless something has changed very recently. And they are very liberal with that free trial. I've spoken with folks from both and they were happy to extend the trial (in my case at least, i.e, education).

You sign up for the trial, download the app and off you go. You can keep the app in both cases to plan/automate your flights after the trial is up. The trial makes available a reduced version of their stitching and modeling software.. which you do lose access to but it gives you an idea of what is available.

I'm not aware of any free software to both collect and process the images (it would depend upon what you wanted to do.. 3D model, orthoimagery, veg indices, etc and what sensor you were using to collect data (RGB camera, IR camera, etc). Just peruse both DD and Pix4D websites and you can learn a lot about what is possible with relatively inexpensive software (if you haven't already). Alternatively there is a lot one could do themselves if they are versed in GIS software like ArcMap, TerrSet, etc. But again because I have academic access these things are 'free' for me (i.e., site licenses) so it's easy for me to suggest something that might not be easily available to you.

Also, if you want to do fairly accurate stitching for orthoimagery it is a good idea to use ground control points which requires high accuracy GPS (sub meter at least, and depending upon the needs of the final product could be much, much lower).

 

[Meta4]

Is there any tool like Drone Deploy, Pix4D etc., that is free to use?

They are all free to use ... for image capture.
They only charge for image processing.

 

[Logger]

Does the Drone Deploy allow images captured while in free mode (expired trial) to subsequently be uploaded and processed during a paid subscription period?

 

[Meta4]

Does the Drone Deploy allow images captured while in free mode (expired trial) to subsequently be uploaded and processed during a paid subscription period?

Try it an see.
I would expect that if you are paying, they won't mind when the images were captured.

 

[bushie]

I use Drone Harmony for capturing of topo mapping. Not free but very reasonably priced. Then use webODM to stitch the photos and produce the surface details - it is free. Finally use either Mapwindow5 or gvSig as GIS packages. Both are free

 

[Meta4]

I use Drone Harmony for capturing of topo mapping.

Drone Harmony charges to use their app to capture images?
The others are free to fly but charge only for using their processing.

 

[bushie]

Drone Harmony charges to use their app to capture images?
The others are free to fly but charge only for using their processing.

Drone harmony has a one off purchase price. It does not process the images, you need other software to do that, hence WebODM