Alex Flies & Maps

I'm going to be upfront here: I made a bunch of mistakes and the first dataset I brought home was basically garbage. So if you're just starting out with drone mapping like I am, maybe this saves you some frustration.

The Mission Planning Part

I used DJI Pilot 2 to set up the grid mission. The property is pretty flat with some tree lines along two sides, which I thought would make things easy. I set my altitude to 100 feet, which in hindsight was probably too low for the area I was covering, and I set my front and side overlap to 70% each. I'd read that 70/70 was a decent starting point for basic mapping so I went with that.

What I didn't think about was how long that was going to take. At 100 feet with 70% overlap on a 12-acre area, the mission ended up being longer than one battery. I only brought two batteries. I had to cut the grid short, come home, swap to my second battery, and then relaunch. When I stitched the photos later in Metashape, you could see exactly where I landed and came back - there was a seam in the orthomosaic that looked terrible and the point cloud got weird in that overlap zone.

Lesson one: plan your batteries before you go out. I should have either flown higher (maybe 200 feet) to cover the ground faster and reduce the photo count, or brought more batteries. I've since started plugging my mission into a calculator first to estimate flight time. There are a few free ones online and DJI Pilot 2 will also give you a rough estimate in the app, but I wasn't paying attention to it.

Ground Control Points - Or My Lack of Them

This is the big one. I didn't use any GCPs on my first attempt. I told myself the RTK GPS on the Mavic 3 Enterprise would be good enough and I didn't want to deal with the hassle of setting targets out. Big mistake, at least for what my neighbor actually needed.

The orthomosaic looked pretty good visually. Colors were decent, the fields were clear, tree lines showed up. But when I tried to pull any measurements out of it - distances, areas - they were off. Not by a crazy amount, but enough that I wouldn't feel comfortable handing this to someone making land decisions based on it.

I went back out a week later and this time I placed six GCP targets. I made my own out of painted plywood, just simple 18-inch crosses painted white and black. I put them around the perimeter of the property and one in the middle. I used a rented GPS unit from a local survey equipment place to get their coordinates - that part cost me about $80 for the day rental which felt like a lot but whatever, it was worth it for the learning experience.

The second dataset was dramatically better. Once I entered the GCP coordinates into Metashape and ran the alignment again, the absolute accuracy tightened up noticeably. I'm not going to pretend I'm producing survey-grade data here - I'm not a licensed surveyor and this was all just for learning - but the improvement was obvious and measurable.

Processing in Metashape

I'm still pretty new to Metashape. I've been using the trial version and I'll probably end up buying it eventually because it seems like the standard tool for this kind of work. The workflow I followed was pretty basic: add photos, align photos (I used medium quality to save time), build dense cloud, build mesh, build orthomosaic. For a 12-acre property I ended up with something like 400 photos on the second flight at 150 feet altitude with 75% front and 70% side overlap.

The processing took a few hours on my laptop, which is not a powerhouse. I've got a mid-range gaming laptop with a decent GPU but not anything crazy. Medium quality alignment took maybe 45 minutes and the dense cloud build took a couple hours. I let it run overnight on the second dataset.

The orthomosaic exported fine and I loaded it into QGIS to have a look. QGIS is free and honestly really powerful once you get the hang of it. I georeferenced the output, poked around with the layer styling, and was able to pull area measurements that matched what I was getting from manual calculation pretty closely. That felt like a real win.

What I'd Do Differently

A few things I'm taking into my next mission:

First, always calculate battery requirements before I leave the house. It sounds obvious but I genuinely just eyeballed it the first time and that cost me a good dataset.

Second, GCPs are worth the effort. Even if the drone has RTK, for anything that needs to be accurate, physical targets and surveyed coordinates make a real difference. I'm also going to look into setting up a permanent GCP or two on my own property so I have a known reference point for practice flights.

Third, bump up the overlap a little. I've been reading that for areas with a lot of vegetation or any terrain variation, going 80/80 is safer. It generates more photos and takes longer to process but the model quality improves and you get fewer holes in the point cloud.

Fourth - and this is something I keep seeing mentioned in the forums - fly in good lighting conditions. I did my first flight around noon on a bright day and the shadows from the tree lines caused some issues with the alignment in those areas. Overcast days are apparently better for mapping because the light is more even. I haven't had a chance to test this yet but I believe it.

Where I'm Going From Here

I want to do more of this. There's something really satisfying about taking a bunch of photos from a drone and turning them into something you can actually measure and analyze. My neighbor was happy enough with the second map that he's sharing it with his surveyor, which I thought was pretty cool even if it's not replacing actual survey work.

I'm thinking about trying a stockpile volume calculation next. There's a gravel operation not too far from here and I know the owner casually - I might reach out and see if I can do a demo flight just for practice. Volume measurement from drone data seems like a really practical application and I want to understand how well it actually works versus traditional methods.

Anyway, if you're just getting into drone mapping, hopefully some of this is useful. I'm definitely not an expert - obviously - but I'm learning and I'll keep posting about what I figure out.