Drone Imagery Collection

3D model, generated from drone imagery collected at 400 ft.

Lately I’ve wanted to get back to testing workflows for collecting aerial imagery by drone — specifically, large orthomosaics, to show on maps.

I had planned to fly in the morning, to collect around the office, but it was not even close to an ideal day for flying — overcast, then once the clouds broke, constant wind speeds of 15–20 mph with gusts up to 33.

So I opted not to fly in proximity to structures in those conditions, but instead to take the opportunity to test the parameters of the platform in a location with lower risk in case of failure.

The Plan

Before heading out to the field, I mapped out the following flight plan:

220 acres | 400 ft altitude | 60% overlap | est. 2 inch resolution

This was estimated to take just under 14 minutes to fly, allowing plenty of time to spare even on a single battery.

Pre-Flight

Once in the field, before beginning the automated flight plan, I first put a battery in and took a quick manual flight to verify my understanding of wind performance and get a feel for how it handled.

The DJI Phantom 3 Pro has an official top speed of 35 mph, and is theoretically capable of maintaining a position lock using a combination of GPS, GLONASS, vision positioning, digital compass, and barometer as long as wind gusts stay below this top speed. Theoretically. Officially, DJI does not recommend flying in any wind over 13–18 mph.

So I flew it up to about 20 ft and flew around a bit to get a feel for the wind. Confirmed GPS hold was working. Took it up to 100 ft, 200, then 400 ft and let it sit there for a couple of minutes. It worked flawlessly, keeping its position within a few feet amidst wind gusts most likely above what I was experiencing on the ground.

Automated Flight and Reality

After the pre-flight test, and after swapping batteries, I again loaded up the 220 acre, 14 minute flight plan, completed final verifications, and hit go.

After an automated takeoff and stabilization, the software flew it up to 400 ft and paused for a few moments before beginning to rotate for the trip to the starting location.

But this part took much longer than I expected. It was sitting up there at 400 ft and seemed to be calculating and recalculating the heading it would need to fly to make it to the starting point. And it began to drift downwind, significantly. The regular DJI software was able to keep a perfect position lock, but the flight plan software was struggling in the wind.

Keeping a finger on the switch to take over manual control if I decided it was drifting too far, I let it go for a while to see if it would pick up and continue. And after about 90 seconds, it did eventually take off on its vector to intersect with the first flight plan row and begin taking pictures.

But by this point, the battery was already down to 65%, and it was clear the plan would not even come close to completing on a single battery.

After flying just a couple of rows, and right after falling below 25% battery level, the software automatically vectored the drone back to its home point and I completed a quick manual landing on a few percentages of battery.

Results & Imagery

After swapping the battery again, I took the opportunity to revise the flight plan, reducing the desired collection from the full 220 down to 110 acres, which would allow for completing the flight plan in increasing wind on the single remaining battery. Take off again, vector to pick up where it left off.

As the flight plan completed and wind gusts continued to increase, landing also became difficult, and manual control was absolutely required, with the automatic return home functionality allowing far too much drift to the point I was afraid of losing orientation and a possible fly-away. Read the DJI forums. It happens.

But, the collection was completed, though with a reduced plan size. And after processing, the results are excellent, with no obvious effects of wind on the imagery. The major takeaway, then, is just that flying in extreme wind requires completely different flight time estimations.

 

Final orthomosaic of new imagery, overlaid on existing basemap.

Conclusion

Ultimately, while it was only a test, the collection was a success. Imagery was captured and processed, I still have my drone, and the results are impressive. As always, the Phantom 3 was nothing but impressive under these conditions. Incredibly solid and resistant to failure.

So it’s clearly possible to fly in these conditions and achieve good results. But it’s anything but ideal. It carries an increased risk of equipment loss or damage, and, if flying around other structures or people, increased risk all around. It requires elevated attention, probable manual overrides, and potentially skill or at least clear thinking to overcome possible loss of orientation at a distance and/or fly-away if automatic procedures begin to fail.

It should not be done in high wind without good reason.

View All