It's been a while since the writing energy has hit. So today has been all about punching through a writeup of doing something useful with imperfect data
#dronemapping #fieldwork #FOSS4G
https://www.spatialised.net/counting-trees-in-sparse-woodland-with-opendronemap-pdal-and-qgis/
@qgis + #DroneDB #microtootorial
You have some drone images over a transect and want to get the approximate area covered [1]
Use #DroneDB CLI (https://docs.dronedb.app/docs/cli) to generate image footprints:
ddb info ./jpg/*.JPG -f geojson --geometry polygon -o footprints.geojson
Drop the footprint geojson into QGIS [2]. Dissolve them (Vector -> Geoprocessing Tools -> Dissolve) [3].
Reproject the result to a CRS in metres (eg UTM) and label with $area [4]
Done!
![[1] Drone images in a transect. The images must have GPS metadata - position and also orientation](https://cdn.masto.host/mapstodonspace/media_attachments/files/114/186/213/932/061/516/original/f352d8bf90b02501.png "[1] Drone images in a transect. The images must have GPS metadata - position and also orientation")
![[2] Image footprints from DroneDB CLI as black outlines. You will need to adjust the styling for this effect](https://cdn.masto.host/mapstodonspace/media_attachments/files/114/186/213/950/846/831/original/916300627ff1a988.png "[2] Image footprints from DroneDB CLI as black outlines. You will need to adjust the styling for this effect")
![[3] Dissolved image footprint polygons as a single purple polygon](https://cdn.masto.host/mapstodonspace/media_attachments/files/114/186/213/653/827/170/original/7a79af9053cf601b.png "[3] Dissolved image footprint polygons as a single purple polygon")
![[4] Dissolved footprints, reprojected to a CRS expressed in metres and labelled with the $area. The label field reads:
to_string(round($area,2)) || ' square metres'
...which rounds the $area attribute to 2 decimal places, converts it to a string, then concatenates it with the context string contaning units.](https://cdn.masto.host/mapstodonspace/media_attachments/files/114/186/213/705/299/604/original/2919095816a95f77.png "[4] Dissolved footprints, reprojected to a CRS expressed in metres and labelled with the $area. The label field reads:
to_string(round($area,2)) || ' square metres'
...which rounds the $area attribute to 2 decimal places, converts it to a string, then concatenates it with the context string contaning units.")
Drone mapping unlocks secrets of 3,000-year-old ‘mega fortress’ in the South Caucasus
A recent study using drone mapping has revealed that Dmanisis Gora, a 3,000-year-old mountainside fortress in the Caucasus Mountains, is far bigger than previously thought. The discoveries also prompted a reassessment of settlement dynamics during the Late Bronze and Early Iron Ages.
More information: https://archaeologymag.com/2025/01/drone-mapping-unlocks-secrets-of-mega-fortress/
Follow @archaeology
here's an #arctic example. Sea ice dtm from imagery flown with an #ANAFIUSA launched off the ice near #FFKronprinsHaakon at 81.5 north, processed in @OpenDroneMap
Colours are deformed bits. Not *quite* right yet but really close.
Part of this underlines the importance of #fieldwork. How do I know when it is "right"? It matches the consensus of people *who were there* seeing with through multiple lenses for multiple goals
I've got a couple of essays / analytical pieces in the works on beautifully imperfect data.
Here's a sneak peek - using point cloud structural analysis to segment ridges in sea ice. I talked about parts of this at ESA's #LPS22, the idea goes way back to 2007 or so - patchy / lack of ongoing work in research makes progress... slow. More soon!
...continuing a little - finding trees in drone data, no AI needed.
Because we have OK-ish representation of what is below canopy, it's possible to try and count things like trees.
A whole bunch of things later, here's a visualisation showing an orange dot for almost every tree.
What's going on with the obvious tree that wasn't picked? Just filtering decisions - if you expect research level rigor from a shakedown flight, #hireme.
Hmm. still a few bushes - however the old mine remains show up wonderfully! I should mention I'm using a cloth sampling filter - reading the original paper to tune it for these data. #researchlyfe
It's so close. The key message is - if you're doing drone things over natural surfaces, *always* check your ground segmentation assumptions, as well as knowing if you're using grass elevation or actual ground ;)
Arrived from hosing trains to a best-yet ground segmentation for Box Ironbark with ground disturbance (old mining features).
The compromise here is some false-negative ground results, in exchange for not identifying as many bushes and plants as ground... I hope...
Ground segmentation is a nontrivial problem for photogrammetric models. There's no cheating with last/only returns, it is all about iterating toward a useable surface morphology.
...a bit of box ironbark. It's not bad - next weather window I'll repeat with double (or more) the imagery. I was after the ground features / mining disturbance under-canopy, and in this forest, it's a fine application for photogrammetry. There's a little work to do on ground segmentation - every zone and target use case has its own tuning quirks!
Spent a little time touching grass - or drone controls - yesterday. Out in Box Ironbark woodland, heavily modified by goldrush mining.
It was a little late in the day for mapping - however, there's a result! I'll add to this thread later.
If you're in the goldfields region and you're after some low carbon aerial imagery / mapping, get in touch! Jobs far from Bendigo might need a little extra planning :D 
Working toward part 3 of a series on drone mapping things, I went looking for vegetation indices in @qgis . I didn't find any so I made some models to compute a few indices:
https://gitlab.com/adamsteer/qgis-things
Hope they help. Feedback welcome - if it means anything other than error correction, send a letter of engagement / purchase order ;). This work is funded by approximately $0 per year at the minute.
A long awaited update to a project first started in 2020 - microhydrology using small drones, @opendronemap, @pointcloudpipe and @qgis
https://www.spatialised.net/mapping-a-small-farm-part-2-microhydrology/
In case you're spending time hunting out how to compute plant health indices from imagery, the @OpenDroneMap
codebase has this wonderful nugget - listing calculations for many established methods (and you can make pull requests to add more!):
https://github.com/OpenDroneMap/WebODM/blob/master/app/api/formulas.py
@sarahdalgulls maybe estimate your average image footprint from camera params and height above ground, then rule-of-thumb it. Say your image is 50m along track, you flew at 2m/sec, your images are 5s apart. You're flying 50m every 25 seconds so in a sequence of images 0-5 you can drop image 1, 2,3 and 4 (this is off the top of my head, image 4 might be needed)
Similar for between lines assuming you flew a grid. Estimate across-track footprint, see what lines can be skipped.
Keen to follow more folk who are involved with #DroneMapping #UAVMapping #Orthomosaics #UAVs
Seeing a number of examples of small drones being used for search and rescue efforts in Syria and Turkey right now. Guessing drone mapping efforts of the destroyed areas will be coming next (or, I hope so).
Nordy Bank is a univallate hillfort on the flanks of Brown Clee Hill, the highest hill in Shropshire.
Here's a composite of some 540+ drone photos, mostly from a P4Pro, but some from a Mavic Pro, all stitched together using OpenDroneMap of the hillfort and it's environs.
OpenDroneMap handled the 2 sets of images from 2 different drones effortlessly!
#HillFortWednesday #Shropshire #Drone #DroneMapping #uk #CleeHills #AerialMapping #DronePhotograhy #HillFort #OpenDroneMap
#Photography has been a natural complement to the #fieldwork that I've performed over the years. It's afforded me opportunities to visit places like this canyon in northern #Iceland. The columnar basalt lining the canyon sidewalls are #geologist's dream!
This view was captured from my #drone during a trip in 2019. I had just finished performing fieldwork focused on #dronemapping several #periglacial and #glacial sites across the island and took the last week of the trip for some sightseeing!
Publication: Improving Approaches to Mapping seagrass within the GBR : https://mdpi.com/1654068 #mdpiremotesensing
#seagrass #seagrasswatch #greatbarrierreef #UAV #remotesensing #dronemapping #drone #earthobserving #machinelearning #spaceborne
