Skip to content

Image transects for ecological assessment

Sometimes we don’t need to create full blown orthophotos and 3D models to understand things. Just seeing an image in a location context is enough. With this in mind, I was recenly tasked by Taungurung Land and Waters Council to collect a set of baseline images along preset transects for ecological assessment. The transects were preset and supplied as line features in a shapefile. After some discussion these were modified to ‘out and back’ flight paths, 20m each side of the planned transect, flown such that the imagery collected in both directions covered the desired transect. Simple enough right?

Flight planning for transects

It sounds straightforward right? Fly lines from A to B and back, capture imagery on the way. However, there are complications – we needed a way to ensure that transects were flown accurately and repeatably. While doing ‘ROV style’ piloting where you set a height and heading and go is a way that works, it’s not good enough, or repeatable enough, for this job. What happens if we need to fly it again? How do we repeat the same lines?

We looked past the button pushing and into the guts of ANAFI flight control, tested, retested, and eventually got results we were happy with. In the end, we constructed a reliable, repeatable flight plan set that took in all the nuances of the ANAFI platform and control system – and delivered the imagery we needed. The general strategy was to fly for 1 – 1.5 cm GSD (40 m above ground), with an airspeed of 2 m/s to limit image motion blur to within a subpixel range.

A note on repeatabilty and accuracy

Repeatability is a catnip word, and can mean vastly different things to different people. Its context here is that we can repeatably collect imagery over the same area every time we fly, and don’t need to make guesses about where to fly. Load up the plan and go!

It doesn’t mean images are taken from the same spot in the air each time, or pixel [x,y] in image flight1day1image10.jpg maps to the same on ground location as pixel [x,1] in image flight3day499image10.jpg with centimetre accuracy.

That level of detail isn’t needed for the project to work – the purpose of this data capture is to provide a set of imagery that fills gaps in time that aren’t accounted for any other way. The methodolgy is fast, relatively low cost, fit for purpose and reliable. And importantly, relatively easy to translate to other platforms. We’d love to keep coming back to grab these image sets, and understand that we might not be selected next time!

In the field

The site was 80% car accessible and also used by the general public as a campground. With this in mind, we didn’t just drive out and try to do it all at once. With some patience and dronepacking, the job was done effectively with lightweight infrastructure and minimal disturbance. For the one site we needed to fly where people were camped and therefore too close to undertake the operation, we got lucky and they decided to go touring for the day. BY the time that flight line came up there was nobody ‘home’.

The field kit. Drone, backup drone, notebook, water, food, first aid, power pack, a nice workspace – all in the 42l Osprey Kamber pack. Designed for ski touring, actually really well suited to dronepacking!

Transects were at the limit of VLOS operation for the ANAFI platform. Because many flights were out and back over water, we were not able to operate from the centre point of the transect – if lake levels drop in future, that will be a preferred option to limit distance from the controller. As a final touch we flew some quick photogrammetric missions to establish lens distortion parameters for each flying day.

Our biggest concern was birdlife – from both a disturbance and raptor attack point of view. Day 1 of flying was closed down by a curious Osprey – and for day 2 we swapped to ANAFI 4K platform, to limit our losses in case a curious raptor got too close. An advantage of the ANAFI platform is folding propellors – these minimise damage to wildlife if it is encountered, although a prop collapse means a downed aircraft. We also fly with care – and program in time to return later if we’re asked by Bunjil (wedge tailed eagle) and friends to clear the airspace.

Data delivery

We delivered 34.7 Gb of imagery and video using:

  • Repeatable transect plans that can be flown again and again as needed for monitoring
  • In situ camera lens distortion parameters, meaning the numbers apply to the situation in the field
  • Lightweight / low impact / ‘leave no trace’ site access methods
  • Low wildlife disturbance
  • A high level of public and airspace safety

Here’s a representation of what we collected – note that the warped imagery is not distortion corrected or color corrected at all – it is straight off the sensor into a projecive transform to image boundaries

Imagery warped to image footprint polygons. The provided transect is the orange line, the blue hatched region shows the polygon around which flights were done – ensuring the desired transect is covered in both directions. Underlying imagery is a VicMaps aerial baselayer. Camera centres are shown as green dots

Flight parameters were validated by generating coverage maps using image footprints. And then checking distances between camera centres to ensure that we flew the program effectively. Image exposure times were generally below 1/480 second, so we know that camera motion at 2 m/s is just a few millimeters – well within pixel dimensions on the ground.

Testing camera centre distances. Orange dots mark camera centres for two images displayed, with their bounding polygons overlaid in purple and green. We expected 20m image spacing, providing 50+ % forward overlap.

The imagery collected isn’t designed for collecting orthopohotography, although over land it works. The water component needs some feature boosting – which is really expected. Using a panorama generator to assemble the imagery with a planar movement between camera centres (eg https://hugin.sourceforge.io/tutorials/scans/en.shtml), and then georeferencing the output, will likely provide data that is just as useable.

We had a quick crack at an orthophoto (OpenDroneMap in fast ortho mode) just to check. Predictably, there were not enough features over water.

Summary

This work falls just as much into our way of work as full on photogrammetric surveys, designing data architectures, processing lidar – and flexed a lot of geospatial muscles. Key components were understanding enough about the client needs to suggest strategies that worked, and then having enough technical depth to complete a data collection program that isn’t a clicking-the-boxes process. And finally, understanding how to collect imagery and video useful for research work. It doesn’t always succeed, however experience gives us our best shot at getting it right.

It was a great couple days of flying and getting to understand the site. And I hope the data contribute to ongoing work to understand how these lake systems evolve and change over seasons and years. If Spatialised is invited back for repeat captures, it’s just a matter of loading up the flight plans again and heading out.

If you’re after similar work – repeatable image collection using lightweight, low impact systems – reach out! We’re happy to chat.

The sales pitch

Spatialised is a fully independent consulting business. Everything you see here is open for you to use and reuse, without ads. WordPress sets a few cookies for statistics aggregation, we use those to see how many visitors we got and how popular things are.

If you find the content here useful to your business or research or billion dollar startup idea, you can support production of ideas and open source geo-recipes via Paypal, hire me to do stuff; or hire me to talk about stuff.