Drone Mapping & Modeling in Pelawatte 101

3D model of the OSC created on DroneDeploy based on data from the April 6th flight (approximately 81 images).

For the last year or so it has been my goal to use drones to better study areas and analyze patterns of land use and vegetation cover. I was first inspired by the work that IWMI was doing to map crops, irrigation and drainage basins using their fixed wing drone (see articles below for references). During the last two years there has been a good deal of excitement of using drones/UAVs for geospatial applications and I am entering the field relatively late. One of the notable highlights of Sri Lanka’s first GIS Users conference held in February 2018 was a presentation on the use of consumer-level drones to map small areas. A Sri Lanka Ministry of Defense Research & Development team demonstrated in real time how easy it was to program a flight path, put a drone up, collect data and then process it so that it could be used for analysis using GIS. Based on their example and the advice of IWMI’s GIS team, I have been working to use the school’s Phantom III Advanced drone to map our campus as well as the nearby Diyasaru wetlands.  This post share some of the results as well as my workflow.

Poster of April 6th drone map of the OSC campus, created on ArcMap 10.5.1.

Results from the May 19th drone flights using PIX4D. This is a composite of two different flights, each with about 30-50 separate images that have been mosaiced and geo-referenced here. If you look closely you will see the OSC Class of 2019 DP Geography class on the west side of the field. Also note the missing large mahogany tree north of  the pool that was regrettably cut down by the neighbors between the two flights.

A Work Flow for OSC

• Step 1  Having a drone is essential and many of the common consumer models can be programed to fly a set flight plan. We used a DJI Phantom III Advanced model that is the older of our two drones (it has already had several major accidents and gone underwater at least twice). The drone needs to be working properly and the micro SD card should be formatted before running the mapping flight. We have been doing a very short test pre-flight on the DJI app that we control the drone with.
• Step 2 You need an app to establish the flight path and program the drone to fly and take pictures at established intervals. We started with a trial version of DroneDeploy and have now loaded in a trial version of PIX4D. Unfortunately, the licensed copies of both of these software bundles are prohibitively expensive for small non-commercial programs like ours and I will have to work out a long-term solution so that students can continue the mapping that we have started. The proposed mapping area needs to be loaded into the drone using the app. For DroneDeploy I made the flight plans on a desktop computer (in the DroneDeploy website) and then imported them onto the phone app that is hooked up to the controller. There are several parameters to pre-set such as the overlap flap area, flight altitude etc.  The higher the altitude, the more accurate and less distorted the stitched imagery but it is coarser (less detailed). For PIX4D I have been setting the flight area on my phone in the field.
• Step 3 At your location you can launch the drone from a cleared area. We use the school field where there is plenty of room and a clear line of sight between the controller and drone (though it is flying by GPS, apparently). When you have got the drone and controller (with a phone interface) unit set up, you are ready for the flight. On DroneDeploy you import the plan, it goes through a number of checks and then asks if you are ready to fly. With the click of a button, the drone hovers and then takes off to run its flight. You can see the images that it takes and the drone’s flight path. Our flights have been set to 75 meters height and they are supposed to be taking approximately 60-90 images for 2 hectares. Unfortunately, we have been having a lot of trouble with good flights but no images being recorded at the end. For this reason, we started using the Trial version of PIX4D and had more success.
• Step 4 After a relatively short flight (5-12 minutes for us) the drone returns and lands and it is time to check to see if the images were collected. If it has gone well, then DroneDeploy will show you a sample mosaiced thumbnail. You need to shut down the unit and then move back to the desktop computer to upload the imagery on to their website where it is mosaiced. It took two hours for them to make the mosaic on the cloud and if you have a license you will also get a 3D model and vegetation map. For PIX4D there is a process where the images are fed from the mini-SD card to the phone and then uploaded online. You can also take the card and load them on to the PIX4D desktop app. All of these steps depend on you having the software and again I am not sure what we are going to do after the trial versions are finished.
• Step 5. In the final step you should have several files to work with. DroneDeploy gives you the orthomosaic and a 3D digital surface model (DSF) as well as a 3D model that you view on the screen. They also provide a KML/KMZ which you can bring into Google Earth (see sample below). PIX4D provides a collection of files (including point clouds, DSFs and more) that can be downloaded. I was most interested in the orthomosaic since I can then pull that into ArcMap where I can use it for presentation and analysis (see above).

 

Getting the DJI Phantom ready to map the OSC campus

Mosaic of April flight overlaid in Google Earth as a KML file.

PIX4D plan as seen on the flight controller.

OSC pool just before a pre-monsoon deluge.

Field and M block before the pre-monsoon deluge.

Drone Deploy plan for April flight

Conclusions

Drone mapping offers a new and dynamic way to visualize landscapes on a relatively large scale. Prices of drones have come down and the challenge is to acquire software that allows you to get your work done. Trial versions of PIX4D and DroneDeploy give you a chance to explore the possibilities and map out a few areas of interest. There are open source options that I need to explore once my trial licenses have run out. Anyone who knows my interests will realize that I am now itching to get drones into forest and mountain landscapes where they can be used to better map vegetation and land cover.

REFERENCES & FURTHER READING

Daniel, Smriti. “The Drone Buzz Over Sri Lanka.”  Sci Dev Net. 19 September 2015. Web.

Lockwood, Ian. GIS Developments at OSC in 2014.” Ian Lockwood Blog. 26 November 2014. Web. (this post features views of the OSC campus via GoogleEarth and ArcGIS).

Mason, Tony. “Put Your Drone to Work. Arc News. Summer 2016. Web.

Siddiqui, Salman. “Sri Lanka’s Drone Pioneers.” ICT Update.  18 April 2016. Web.