Drones at NBR: software

This is the sister post to the one on Drones at NBR: hardware

There are two parts to this post: Things we use, and things we’ve considered.

Things we use (plus a comment or too):

• Creation and execution of mission for RGB missions (the vast majority):
  • Pix4DCapture.
    • Super easy and smooth. Works really well for us where we can afford to be a little rough regarding mission coverage (our missions are small and we have lots of batteries) and also need to build the mission there in the field during the same visit that we’ll fly. Everything happens on the phone. All you really need to do is; start a new ‘project’, walk around the trial site to adjust the bounds of the mission correctly, check one or two settings, save, connect, fly. When you come back to that site next time, you have a ‘project’ and a ‘plan’ (i.e. a site and a mission) ready to load and go again.
    • I can’t remember exactly how it goes with sharing actual mission files (I think it was: share the ‘Project’ folder with all files in it, but in the ‘Mission’ sub-folder you only need the ‘details’ file, maybe the .p4d file too), but I’m certain it does allow for KML files to be loaded. This means that if you work in a team, it’s pretty easy to ensure everyone has access to files that allow them to fly a mission without re-mapping the site (i.e. save 15 minutes). We draw the KLM files after the first mission at a site is flown, then these files are loaded to DropBox, which everyone who is flying the drones has direct access to on their phones (Dropsync with the Pro version purchased is the app we use to make the KLM-to-phone process automatic).
    • I still like the DJI Go 4 app better for flying ‘free-flights’. This makes for a little fidgeting when swapping between a mission and free flight. When I fly with Pix4DCapture, I find I need to have it as the default Phantom app. The same with DJI Go 4. Swapping this setting means turning on the drone, turning off the default app that automatically starts, opening the other, selecting ‘yes’ in the pop-up that asks if you want to use this app as default, then restarting everything so that the new default app actually connects to the drone. It really only takes 30 seconds, and there is probably a better way, but I haven’t found it.
    • From chatting with colleagues, the app works better on Android than on Apple, but it’s still really good on Apple.

• RGB image processing:
  • SOLVI. The main advantage of SOLVI is that it’s cloud based, so frees up time and space on desktops. We use it to stitch together orthomosaics, build plot grids (see next point), and some basic analysis, and it’s doing a pretty good job. The service is cheap, quick, reliable. It might not have the range of analyses as some others that focus elsewhere than agriculture, but it give us what we want. You can read more about it in a blog post they did with us. Just to be clear, we pay SOLVI for their service  – we don’t get any commercial advantage from this blog post.
  • SOLVI has also developing a plot identification systems. This is super important for us if we want to be able to use remote imaging for real research data as opposed just a record of how things look during the year. Their automatic system seems to work well in cereal crop plots, but it’s pretty useless in sugar beets. Fortunately, the more manual system is pretty smooth – generally it takes about 2 minutes to get the plots laid out right.

• Trial site location files:
  • We also use the SOLVI zonal statistics grid tool to create SHP files of the boundaries of trial sites. We then use QGIS to turn these plot boundaries into a combined trial polygon. It’s usually saved as a KML files. These KML files are uploaded to the shared Google map of trial site locations, and can be saved to mobiles and used in mission planning in Pix4DCapture. HOWEVER, to do this, it is really important that the corners of the trial sites can easily be seen in the images on SOLVI. Some sort of marker is needed. I highly recommend the purchase of a meter or so of chequered lino (rutiga vinylgolv). It’s light, durable, you can cut it into a 2×2 square with scissors, and it’s cheap: we got 10 markers for around 250:- (€25). It was 1 meter of 4 meter wide lino from Bauhaus.

• Creation and execution of mission for thermal missions:

This used to be in “THE OLD WAY” for the RGB missions. Why? Before I had built experience with our system, I was really worried about drone battery life and the storage of images. So, I was very keen of having tightly planned missions that did only what I needed and no more. To achieve this, I was convinced we needed to do as much pre-planning in the office that we could. Thankfully, we don’t have unmanageable issues with battery life or storage, so we can use the Pix4D Capture system for RBGs. But then we got the thermal drone, and the following is the only way we’ve managed to to be able to run a mission with both the RGB and thermal camera running…

• • Creation and execution of mission (thermal):
    • DJI Flight Planner (djiflightplanner.com). Not as many options as other programs like Mission Planner, but just the right level of control for those using DJI and just needing to turn accurate polygons into standard grid missions. Has a strange mix of imperial (height in feet) and metric (speed in km/h), but they’re easy to work around. Worth the USD99, once the 30 day free trial has expired.

• • Uploading/ Checking/ Modifying/ Saving missions (thermal):
    • Litchi Mission Hub (flylitchi.com/hub): Again, pretty user friendly when using in a web browser. Free to use. Haven’t tested how many missions we can save yet. The best thing with Litchi Mission Hub is that it syncs really neatly with the mobile app, so you desk-top missions are ready to fly directly once you’re out in the field.

• • Executing missions (thermal):
    • Litchi app. Have been using it on Apple. Had a confusing default setting of auto-record (so took video instead of photos on the first attempt), and on my little iPhone 5, I can’t see the return home button that’s meant to appear after a mission has completed. It’s easy to make modifications to a mission in the field and is generally pretty straight forward to use. Not sure how it goes with Android yet. For the 250SEK or so it cost, and given the way it seems to sync seamlessly with the Litchi Mission Hub, a good investment for our situation.

• Connecting remote imaging data to other trial data.
  • NBR uses Excel a lot. To link it and the plot level Shape files from SOLVI, we use a combination of… VBA, R, Python, and QGIS.
    • VBA take the raw data and necessary parameters for mapping the data out of Excel and passes it to R.
    • R joins the raw data and the Shape file, writes a Python script that QGIS will automatically seek and run on start-up, and starts up QGIS
    • QGIS visualises all the data in a user friendly interface.
  • There are still a few bugs in the system, and steps that could be more automatic, but it’s smooth enough to use without causing unnecessary stress.

And the things we’ve considered but aren’t using (plus an explanation of why/ why not):

• Trial site locations
  • Nothing, actually. The ArcGIS suite was known and available from the start, and provides everything we need here.

• Creation of missions
  • I used Mission Planner previously, when I was flying 3DR Solo drones. And I liked it: it gives a lot of options and control. It’s also free. Unfortunately, as far as I can tell, it doesn’t work with a DJI as well as other options: the DJI Phantom camera is not a listed options, and the you don’t get any output from the program that can be used directly in with the DJI (I haven’t looked too deeply at this point). I was able to create missions in Mission Planner that are very similar to those made with DJI Mission Planner (using the camera parameters for a 16:9 aspect; focal length 4.175, sensor height 3,5156, sensor width 6,25, image height 3078, image width 5472). I was also able to easily take these over to the Litchi app by copying the flight details into an Excel file and running a bit of VBA, but that was fraught with danger, and clearly not as simple as with DJI Flight Planner. But it’s a good (and free) back-up plan.

• Execution of missions
  • MicaSense Altlas Flight: super easy mission planning, but you can’t run the DJI RGB camera from it.
  • DJI Pro 4: need to buy a ipad mini, which we’re not too keen on given everything we do is Android.
  • Senteras FieldAgent: couldn’t find it.

• Image processing
  • PrecisionHawk: I’ve got one of their free accounts, which gives 5 free orthomosaics, but we need way more than that. They have a pretty impressive range of analyses available, but many of these require a multispec camera – which we aren’t running – and the subscription fee of US$3,500 per year was just not competitive. If we expand our use of remote sensing in research, we’ll definitely be revisiting them.
  • Pix4D: comes pretty highly recommended for agricultural applications, but again and just for now, they have too much on offer at a price that is too high. We’ll also revisit them if we expand our application.
  • Agisoft: I really like the online image processing options as they free one up to concentrate on other things while the images are processed. But they also restrict the level of control one has. At this stage, it also sits in the ‘too much on offer/ costs too much’ basket, and will be revisited in time.
  • hiphen: Looks very good. Agriculture focus. Have research sub-focus, so can do plots (PhenoScale system). Applies BrAPI / MIAPPE standards. Nice people (I met them at MODEL-IT 2023)! If I remember correctly, not everything was automated, but the major drawback was the cost.
  • agremo: No automatic orthomosic creation. Quite a lot of analysis tools. Nice people.