Crunching Drone Data: How Propeller Handles Complex Datasets

At Propeller, we say we offer “quality-assured data processing” and we mean it. We’re proud of this fact, and we think it sets us apart from the competition.

“But do you actually have real photogrammetry experts checking my drone data, every time I upload?” skeptical customers ask. The answer is “Yes. Yes we do.” Every dataset with GCPs uploaded to our platform is validated by one of our in-house data specialists; a bunch of clever folk whose job it is to ensure that ground control has been accurately applied to aerial images, and quality processing can happen.

In the process of doing this, our team sometimes notices bugs or anomalies. In these situations, they apply their specialist knowledge (and, often, many many hours of research) to figure out exactly what’s gone wrong, and work with the customer to set things right.

Here are just some of the problems our data specialists have solved for customers:

 

Exotic coordinate systems

If you’ve ever delved into the world of geospatial analysis, you’ll have run into the seemingly endless array of coordinate systems used across the globe.

While some of these coordinate systems exist in large repositories maintained by groups like the European Petroleum Survey Group (EPSG), there are thousands more that exist only in local government repositories, or a surveyor’s database.

Fortunately, Propeller supports the defining of new coordinate systems (as long as you know the right parameters to describe it). One example is the Coorow94 projected system that covers a sparsely populated area in Western Australia. This system has no corresponding EPSG code, but does have a full suite of projection parameters describing its relation to the GDA94 datum. Using these numbers, our data specialists were able to set up this coordinate system for use in future surveys.

Propeller plaform working with drone data in exotic coordinate systems COOROW94Propeller plaform working with drone data in exotic coordinate systems

There are times when customers don’t want a whole new coordinate system defined—just an existing system in different units. In these cases, Propeller can convert outputs on request to support units that might not exist for certain published projections.

All these various coordinate systems can create a reasonable amount of confusion! So if you’re ever unsure about what coordinate system to use, or which EPSG code is correct, it’s good to know you can reach out to our expert team (via support@propelleraero.com).

 

Multiple camera models

Some jobs require more than one drone to complete. Surveys of large areas, easily covered by a fixed wing, may also include smaller vertical assets better suited to being flown with a multirotor drone. Naturally, each drone comes with its own hardware, including the camera, which can lead to inconsistencies in your dataset.

Propeller’s expert team can fix this problem by generating individual camera models for every camera used in a dataset, giving our customers the confidence to use whatever tools they need to get the job done.

 

Datasets with different flying altitudes

Some jobs require multiple flights and, as a result, the elevation data attached to different images can vary wildly. This causes problems during processing.

Altitude variance can be caused by a number of things: incorrect drone GPS settings, not taking into account the takeoff location altitude, or a firmware issue which, if not corrected for, can leave your data looking like this:

Solving altitude difference challenge in drone data

The example above shows building inspection photos taken on the same site. One group of images with an altitude of 80m above sea level, and another group 49m below sea level, both captured on the same day using the same drone. Without adjustment, photos are able to be stitched together, but two separate models are generated.

After our data specialists corrected the negative altitude issue, the model looked much better.

Corrected 3D model of a building inspection

Similar altitude issues can crop up in the case of large area surveys, when drone pilots need to stop and change batteries. With many drones that tag photo elevations based off barometer readings, these kinds of variances are almost unavoidable, Thankfully this is something our team is very familiar with, and can handle without fuss.

The example below shows a large area surveyed with multiple drone flights, where one part has a 20m altitude difference set at the takeoff location.

Distorted 3D model due to altitude variation

Again, after our team’s realignment and processing adjustments, the model generates correctly.

Corrected altitude variation in drone data

 

Very large data sets

Do you want to survey an area four times larger than Manhattan? No worries. Our expert team can work with you to get your very large datasets online.

The largest pixel count we’ve processed so far was submitted by Ninox Robotics. This survey in Western Australia contained 10,500 images from 12 flights, over 35 Mp each. The total area spanned more than 246 square kilometers, and the reconstructed point cloud included more than 3.5 billion points.

The final orthomosaic file weighed in at 200 Gigabytes (impossible for an average PC to load by itself, yet easily viewable using Propeller’s secure cloud-based platform).

Note: Oversize jobs (greater than 1,000 photos) fall outside our standard pricing model, so if you’re working on a job of this scale, it’s a good idea to get in touch with our team. We’ll figure out the best way to deliver this data to meet your needs.

 

Bringing in design files, overlays and historical data

When customers are upgrading their mapping workflows to include drones, the question often arises, “What do I do with my old data?”
Luckily, Propeller supports the uploading of preprocessed data, from design files to vector overlays and orthomosaics. By default, we accept KMZ data for overlays and GeoTIFFs for orthophotos and elevation models, each of which is checked and validated by our data specialists before being brought into the platform.

Cross section compared against design surface

Have other formats? If you have DXF or DWG CAD files and aren’t sure how to convert your data, simply reach out to our support team and we’ll help you get your models onto the platform to be used for volume calculations and more.

Note: Files that don’t explicitly state which coordinate system they have been generated in can be tricky, so it’s always best to provide the coordinate system and unit of measurement of a new file at the point of upload.

Design files overlay on drone data

Conclusion

While we’re always trying support our customers to upload good quality data (as this optimizes turnaround time), we understand that every dataset is unique. Some jobs by their very nature require more complex solutions, and sometimes things simply go wrong in the field.

Whatever the challenge, our team is quite simply the best in the business when it comes to processing challenging datasets and working with our customers one-on-one to ensure low-friction processing in the future.

Interested in using Propeller solutions for your data? Get in touch with our team.

 

 

You might also like:
Using Local Grid Coordinates on Surveys with Propeller
When Sky is Not the Limit: Largest Datasets Processed by Propeller
How You Actually Get Drone Data Accuracy Down to a 10th of a Foot

Get blog updates directly into your inbox

Be the first to know about Propeller's news and features