Surveyor of some 40 years and teacher at the Canberra Institute of Technology, Noel Burton of Detail Survey, addressed our 12d 2018 Technical Forum audience about using drones for surveying work.
Over his career, Noel said, he’s seen things progress from a non-technology type surveying right through to where we are at the moment, where he’s happily going out to use drones for most of his survey work.
The main focus of his talk was on using drones for those who don’t necessarily do so every day in the industry, to give a broad overview of what they can do.
What is a Drone?
A drone is currently known as a Remotely Piloted Aircraft (RPA).
Broadly speaking, there are two types of RPA – Fixed Wing and Multi Rotor. Fixed Wing RPAs fly like a plane whereas Multi Rotors will generally hover when you take your finger off the control.
Fixed Wings are the original RPAs used in surveying; they are traditionally launched from a catapult or manually propelled (thrown up in the air). These come with some issues, though – they require room to land, they’re quite expensive…and they’re apparently the favourite food of Wedge Tailed Eagles! Countless videos exist of these majestic birds taking exception to Fixed Wing RPAs, and while it’s quite funny to watch, it can also be a very expensive loss for anyone involved!
Multi Rotors, on the other hand, have been adopted more recently by surveyors, including Noel. Their popularity is down to a number of factors. They can take off and land vertically as well as hover, and they’re good not just for mapping but also for video and photo monitoring of structures such as power lines. They’re also quite inexpensive and easy to use, and their resolution and quality actually compare well even against some of the really expensive Fixed Wing RPAs. Oh, and as a sideline bonus, eagles, hawks and crows seem only amused by this strange ‘bird’ in their skies.
Rules for Using RPAs
The Civil Aviation Safety Authority (CASA) is the governing body for anything that flies in Australia. In the last couple of years, their rules have softened slightly. For instance, with Noel’s drone he can happily fly it and charge money for his services under their rules.
Their website lists the rules for flying RPAs – most are just common sense , but it’s important to know them anyway – Noel took us through some of the important ones. They cover things like only flying during the day, keeping the drone in your line of sight, not flying it above 120m, keeping it 30m away from other people, and never flying it near a bushfire zone. If a RPA is more than 100g in weight (which they pretty much all are), it needs to stay away from airports.
If you’re licenced as a drone pilot, you can apply to break some of the rules slightly. The corollary of this is that if you follow all the rules, you don’t need a licence according to CASA. However, Noel suggested talking to an insurance company as there could be liability issues if someone is injured by a drone, etc.
Noel said that the first step to mission planning is to check whether you can fly a drone in your selected area – Google search ‘can I fly there?’ There are also phone apps available to assist with this. The result, either way, is that users can be notified of any restrictions in selected areas through use of an interactive map.
Noel’s next step is to get a surround. With a lot of the apps for flight planning, if you can create/receive a KML file of the extents of the proposed area for survey, you can use that as part of your flight planning. To get these, either use GoogleEarth yourself or get the file from the client.
After that, you can plan the locations for your Ground Control Points (GCPs) – i.e. the known points for being able to position the drone survey.
Some of the ‘fancier’ drones have accurate positioning in them so you don’t need GCPs, but with the more basic models, it is necessary to enter targets.
The spacing of the grid depends on how high you’re going to fly – if you’re going to fly high, you can have the lines further apart; if you’re flying lower to the ground, keep them closer together.
Once you’ve generated the KML file, we can bring it into an app that will run on the computer (or even an iPad). There are a few apps out there – Noel uses DroneDeploy. It does all the processing for you if you want to submit online; if you don’t want to do that it will design the flight plans for you and you can do the processing yourself (as Noel does – this method is free).
With this software, you can enter the height you want to fly, the forward and side overlap (sufficient overlap is necessary to create the 3D model), and the flight line directions.
Once you’ve generated the flight lines for the mission, you can calculate with the software how long the mission will take, the area you’ll cover, the number of photos, and the number of batteries you’ll use.
Noel showed footage of a mission that was completed using an iPad Mini and then simulated this mission live on screen for the audience, pointing out where the drone started taking the photos, and then what to do with those images/information.
Processing Data Gathered by RPAs
There are two main brands of processing software – Pix4D and Agisoft PhotoScan (now called Metashape). Noel uses Agisoft, which he says is fairly easy to use.
Agisoft – steps:
- Adding photos
- Deleting unwanted photos
- Setting coordinate system
- Aligning photos
- Importing and referencing GCPs
- Creating the point cloud
- Filtering unwanted points
- Creating mesh and imagery
Adding and filtering photos
The first step is to read in the photos taken by the drone and select the folder, then delete the ones you don’t want. You can then estimate the image quality – the program will go through to find the ones that aren’t high enough quality and get rid of those.
Aligning photos and referencing GCPs
The photos are in latitude, longitude and height, and we convert them to MGA coordinates (the local coordinates). This is useful because when you survey the points they’ll use MGA and you need to be able to import them into the actual job to be able to manipulate the images. Also, there’s an option to align photos – the program will try to match pixels on the photos to build a model so you can read in your GCPs and identify them on the photos. It’s just a simple CSV file. Noel just uses regular black and white tiles for this – positions each on the ground as reference points for the photos. This gives you an idea of residuals to increase accuracy.
Creating and filtering the Point Cloud
When the initial point cloud is generated, it includes everything; running the Classify Points option allows you to filter out items such as trees, which are placed in different layers.
Creating mesh and imagery
We generate the mesh from the point cloud (just the ground layers). When we get the triangles, we can then use Decimate Mesh to filter out areas that are within a certain tolerance. From there it’s just a matter of getting out the rest of the processing. Each day, Noel gets things to a point where he can run this option overnight – batch process to build up texture, create imagery, export points (e.g. LAS), create a GeoTIFF file for accurate imagery, generate reports, etc.
Noel showed a KMZ file that came out of the job he was demonstrating, commenting that he finds using a drone for a job so much simpler and quicker than previous methods – he greatly appreciates this technology for saving footwork and increasing efficiency. The work has lots of applications. Noel concluded by stating that every Surveyor should have a drone on the back of the car – it is such an important tool these days.