Hi all
First time posting here and look forward to reading other posts.
After having spent 40 years in aviation and now beef farming for the past 4 years in Australia I have been researching the idea of combining the experiences into a business concentrating on providing a no fuss multispectral ag service for farmers in the local area. The area I live in (Gippsland) is a very fertile farming district with abundant rain fall that grows a variety of pastures, crops and vines so the opportunities are many and varied.
I have been doing a lot of research and finally coming up with a solution that i think will work well. Part of this solution includes using ODM to process my indexed orthomosaics.
I am trying to be financially efficient and keep costs down but understand I need the right tools for the job which means upgrading computers and alike.
I have a couple of questions for the forum.
I only intend to output 2d indexed maps but have heard that ODM can be quite memory intensive when processing data. I intend to buy a MacBookAir with 16GB RAM (Max allowed) is that sufficient? I know more is better but I am trying to keep initial costs down.
As far as mapping solutions go I have been looking at Litchi but note that it does not have a mapping capability. Is there a set of tables or an input form that is able to provide a flight path and spacings for a given height, speed, overlap, FOV, Focal length etc etc. I’m happy to manually construct the mission. I know there is other software eg Pix4D, that can create the whole flight but my research seems to indicate that there are some reliability issues with these programs Litchi being very reliable.
ODM was recommended to me by another user and I know it is open source which does worry me somewhat from the point of view that I have limited computing skills when it comes to interfacing at the operating system level if you get my drift. I would like to know what other people experiences are in particular ease of use when producing eg NDVI maps etc I guess what I am asking how intuitive is ODM to use. Of course I will read as much as possible but that doesn’t always cut it.
Thanks and Cheers
BW
1 Like
First of all, welcome to the forum! Incoming wall of text, don’t be scared, all of it hopefully helpful.
Don’t worry about command line interaction! The WebODM gui is very user-friendly. You can look more in the documentation, but basically ODM is the core photogrammetry engine that stuff like WebODM is built on top of, so you don’t lose functionality.
In reference to buying a computer for photogrammetry purposes. Basically more RAM = larger maps you can process and a faster CPU with more cores = faster processing times (but doesn’t limit you like RAM does). 16gb of RAM will likely process medium-largish maps, although it’s sometimes hard to tell. Actually some our of community members have recently started trying to answer this very question over at https://www.opendronemap.org/odm-data-benchmarks/ . Something I immediately see might be a problem with a Macbook Air is potential overheating if you intend to do any work in non-air conditioned environments. Macbook Airs aren’t particularly known for good airflow. Perhaps consider a Macbook Pro if you really want to stick with Mac and maybe one of those cooling mats, or at least an elevated stand if you want to extend it’s life by avoiding overheating during long map processes (https://www.amazon.com/dp/B00WRDS8H0/ref=cm_sw_em_r_mt_dp_U_TawMEb58PYCSW).
Thirdly, a couple of questions that will help narrow down how I can help:
- In regard to flight planning software, what drone(s) will you be using to fly these maps?
- How large are the areas you intend to fly?
Since I assume you will be mapping crops/pasture what are some common target species you may be focusing on?
Obviously I love ODM, but also consider trialing Pix4Dfields (costs $ subsciption and is a different product than Pix4Dmapper). It is an awesome product for farmers and uses a different approach to photogrammetry that trades accuracy for speed. I’ve literally had maps process in the field, it’s pretty cool for looking at large problem areas when centimeters don’t matter.
P.S.
If you don’t mind me asking, what ag multispectral services/deliverables you will be providing? I’m always interested in how people are using this recently democratized technology!
2 Likes
Hey @barneyw welcome to OpenDroneMap. I’m based in Benambra, licensed to pilot rpa, lots of experience with aerial imagery crunching/research and already have a data processing setup. Drop me a line 
3 Likes
Welcome! Some quick thoughts:
- Best value per dollar will always be a custom build desktop. There are tons of resources to help (PCPartPicker, for instance), or companies that charge minimal markup of about 30% for doing the work for you (CyberPowerPC).
- Go for 32gb RAM at least, and look towards CPUs which balance strong single-thread performance with number of threads. This will be a push/pull to optimize, unfortunately.
- Decide what sUAS you will use before you look at flight planning software. I choose to use ArduPilot as my flight stack, so I can plan with pretty much anything I want from Mission Planner to APMPlanner2 to Tower to Solex to QgroundControl to UGCS.
- A bit of a nitpick, but you can’t really call a product NDVI unless the sensor you’re using actually has the proper bands and band cutoffs to match the canonical definition of NDVI. Something which lets near Infrared pass isn’t really NDVI, but you can generate NDVI-like indices which accomplish the same task of showing relative vegetative vigor.
- Choose a sensor that makes your job easier. Off the top of my head, I like the Parrot/Micasense offerings, especially since they have integrated daylight sensors for easy calibration so you get repeatable results. They cost a penny, so I didn’t use them, but instead used a converted GoPro with a Blue/NIR filter.
- WebODM Lightning is a fantastic platform, and offers an ease of use closest to Pix4D in my opinion.
- I used QGIS for all my data analysis and report generation, whether I processed my data with Agisoft, Pix4D, WebODM Lightning, or Microsoft Image Composite Editor.
1 Like
Rainy Rockies appreciate the reply and have taken on board what you have written. I intend to provide a simplified ag service to local farmers and have a number of contacts to get me started. Interestingly enough I had a chat with a person who sells drones etc, and like anything it’s very easy to get caught up in the hype, and he brought me back down to earth a little but suggesting that a lot can be achieved with RGB mosaic (VARI) as most farmers can become a little confused and moreover, if you can’t explain it to the farmer in simple terms then all is lost. So that’s a starting point. This why I intend to ensure that I am well across the technology and have been learning all about how colour and colour filters work and will be doing the necessary theory before venturing into more advanced analyses NDVI, ENDVI etc. Having said that I need to ensure that I provide useful data for farmers. Where I live we have a lot of crops including potatoes, legumes and vines not to mention the dairy industry that is always sowing a variety of crops such as turnips and chicory etc and trying to become more efficient.
Cheers
Saijin Thanks for your response. I am setting up a P4P capable carrying up to two MAPIR Survey 3 camera. I will have three of these camera in the following filters OCN NGB and RE. The owner of a production company who has been Providing ag services for quite some time uses one of these camera on a Mavic Pro mounted on the front with great success. with my setup I will be able to capture RGB + any one or two of the other bands the only real limitation is flight performance but it’s a good compromise producing good results. Thanks for all the other info I will check it out. Cheers
2 Likes
Hey Adam
Appreciate the offer - how would you like to get in touch. Is there a way I can privately contact you through this forum.
Cheers
Tony
I sent a message click your user icon (top right) to see notifications.
Cheers,
The MAPIR cameras are great, and they have some great documentation available to help you learn how to do sUAS aerial survey and processing as well. Just make sure you really think about whether or not you should use the reflectance calibration target with your workflow, and if not, how you’ll address that variability in-season with the growers.
Nothing wrong with a P4P. I run Solo
Hi.
I have trying to scout the web about info regarding converted GoPro and the possibility to use it for plant health but don’t quite find some useful.
Can some of you share some knowledge around this ?
What I had in mind is to buy used GoPro 4+ and new lens without blocking. Is there some more one have think of ? (ir filter and so on)
The reason why 4 is that it’s the last GoPro with m12x0.5 thread. 5 (who has GPS) and above is having m12x0.35 and those lenses X10 more expensive.
(I’m aware that it means that I need to geo reference some of the images to be able to have as overlay on Google maps).
Every input is welcomed.
1 Like
I use a converted GoPro Hero4 Black with a full spectrum conversion from Ilija at KolariVision Infrared. I used his Blue/NIR NDVI filter. He has other filters as well.
It gives you results that correlate highly with NDVI, but nothing can be actually NDVI without the same bands and cutoffs.
I used it for two years doing Agricultural Field Survey.
1 Like
Hi Saijin. .
Is it possible for you to upload some example from the GoPro 4 to my webodm server ?
Would be really helpful to see what this converted GoPro can “bring to the table” sort of speak.
1 Like
Ok … checking …
Hmm… I’m don’t quite get what I’m seeing here … beside the pink color that must be orgianted from the nonexistent of uv blocking coating of the lens.
I pick a screen grab off just random test with rgb.
Can you give some input in your workflow regarding multispectrum and plant health ?
Is there something that this gopro4 without uv blocking gives you instead from just rgb ?
** edit **
Found a paper that involves NDVI among others…
** edit **
You can’t use NDVI with that setup, since it isn’t an NDVI sensor/filter array. You need to use a custom index, though ENVDI is pretty acceptable.
Also, the pink color is the Near-Infrared saturating the Red band, and is pretty much normal behavior for CMOS senors.
UV will mostly register as excessive noise on the Blue band, but this filter does in fact filter out UV.
Seems that I have tons of reading and testing to get a grip around this.
Just one final thought before I press buy on the used gopro 4 black I have found on a auction site.
Do you mess with manual geo reference a few images by adding geo based from google maps like location for a house / crossroad and so on. Or do you leave as is (webodm will then not place your survey on the map.)
For the lens I’m thinking of buying from ebay, from what I understand blue nv filter is a good idea to filter out noise ?
Something else to think about ?
(…when that china parcel final arrives, I certainly would have lost the interest anyway, so perhaps it matter whatever I choose in lens/filter and so on … 
)
1 Like
I georeference my images using the GPS data from the 3DR Solo that carries it and issues the photo command.
You can manually reference, but it is a bit of work. If you’re going to be on foot, you can record a GPX track log with your phone and USE GeoSetter to tag the images.
What specific filter are you looking at?
