Processing drone photos general questions

#1 · August 10, 2024, 10:04 am

I had been thinking about getting the remote pilot certification for a while in order to be able to capture imagery that is more current and better resolution than data available online and it would be my actual work product to incorporate into anything I was doing. So, I went ahead and got certification, have my drone registered and insured, and have added it into business insurance policy. I'm currently running some practice flights to get comfortable with balancing flight altitude / desired resolution and processing the collected data.

There are several software options for processing so I haven't determined what I'll ultimately end up using. It looks like Global Mapper may have that ability, so I might use that as I have been using GM for decades (I'll most likely have a go at GM this afternoon). But for now, I have been following some online tutorials using an open source program called WebODM. I probably need to work on getting the processing settings setup correctly as I was expecting better results.

Questions:

1) What software are you using? Pros and cons would be great to hear about.

2) What are some tips you might share with a newbie on getting good results (both field practices and processing)?

3) If anyone is using Global Mapper for this, would love to hear feedback about the process through GM.

#2 · August 10, 2024, 7:37 pm

You should research Aggiesoft. It is a sophisticated imagery processor. I'm not well versed enough to flesh out its capabilities relative to other software though.

#3 · August 12, 2024, 7:40 am

Quote from murphy on August 12, 2024, 7:40 am
I've used ESRI's SiteScan, LP360 and TBC-Enterprise but have never touched Global Mapper. I have noticed a difference in the software's ability to handle photo data with varying side overlaps. SiteScan can take my 50% side overlap imagery, taken simultaneously with lidar, and generate decent orthos. The other software will show distortion or bizarre pixilation on elevated features, e.g. houses conveyors, some trees. This will likely be a non issue for you since it's standard to fly a 75% to 80% side overlap on a photo only mission.

The fieldwork can be a fun challenge. Start with an acceptance that there are a large number of variables which will make it difficult to develop immutable rules in the short term. Insolation, wind speed and variability, height AGL, IMU errors, camera specs, shadows, surface homogeneity, flight speed and other factors must all be considered. Also, what works best for an orthoimage may not be best for generating contours. Since you mentioned that your preference is the former, the following considerations will lean that direction

1. Create a kml in GoogleEarth of your flight area of interest. Square it off and worry less about total acreage and more about flight paths and where your UAS will be turning on its approach to another flight line. If your flight controller permits, import the kml to establish your flight AOI. Email a copy to yourself and open it in GE on your smartphone to assist with the placement of panels once onsite.

2. Where you place the panels is often more important than the quantity, with caveats. Start with panels about twenty-five to fifty-feet in from the four corners of your AOI plus one in the center. Similar to providing construction control for machine modeling, interpolation is better than extrapolation. However, you have to have a small amount of extrapolation because you're stitching pictures together and the panels need to be in the photos. If you're generating contours, place additional panels on the highest and lowest points. With ortho only, you can put the panels anywhere that's open. If you're generating contours, don't place them within ten feet of any grade break. I look for 20x20ft grade planes, then set my panel in the center. Flat is best but a slight slope is okay too.

3. Lighting is a major factor if you're trying to get a good pic. Overcast days tend to be good. The long shadows in the morning and evening take away from the usability of the image. Flying midday eliminates shadows but may degrade photos of concrete or light colored and reflective surfaces.

4. Don't skimp on checks. Fill in the gaps between your panels with red or similar colored painted dots. You can also use red plastic dinner plates with a nail in the center or print your own on mylar.

5. If you start generating contours, understand and use the 2023 ASPRS Positional Accuracy Standards and give due respect to the science of photogrammetry (I'm sure you will, but there are many sUAS Airmen who seem to think they know more than photogrammetrists or the FAA and can pick and choose which parts of the guidelines to follow).

I suspect that after a dozen or so flights you'll find a simple and quick workflow. I've stopped counting how many times measuring from my own ortho has saved an additional trip to the site.

Good luck

I've used ESRI's SiteScan, LP360 and TBC-Enterprise but have never touched Global Mapper. I have noticed a difference in the software's ability to handle photo data with varying side overlaps. SiteScan can take my 50% side overlap imagery, taken simultaneously with lidar, and generate decent orthos. The other software will show distortion or bizarre pixilation on elevated features, e.g. houses conveyors, some trees. This will likely be a non issue for you since it's standard to fly a 75% to 80% side overlap on a photo only mission.

The fieldwork can be a fun challenge. Start with an acceptance that there are a large number of variables which will make it difficult to develop immutable rules in the short term. Insolation, wind speed and variability, height AGL, IMU errors, camera specs, shadows, surface homogeneity, flight speed and other factors must all be considered. Also, what works best for an orthoimage may not be best for generating contours. Since you mentioned that your preference is the former, the following considerations will lean that direction

1. Create a kml in GoogleEarth of your flight area of interest. Square it off and worry less about total acreage and more about flight paths and where your UAS will be turning on its approach to another flight line. If your flight controller permits, import the kml to establish your flight AOI. Email a copy to yourself and open it in GE on your smartphone to assist with the placement of panels once onsite.

2. Where you place the panels is often more important than the quantity, with caveats. Start with panels about twenty-five to fifty-feet in from the four corners of your AOI plus one in the center. Similar to providing construction control for machine modeling, interpolation is better than extrapolation. However, you have to have a small amount of extrapolation because you're stitching pictures together and the panels need to be in the photos. If you're generating contours, place additional panels on the highest and lowest points. With ortho only, you can put the panels anywhere that's open. If you're generating contours, don't place them within ten feet of any grade break. I look for 20x20ft grade planes, then set my panel in the center. Flat is best but a slight slope is okay too.

3. Lighting is a major factor if you're trying to get a good pic. Overcast days tend to be good. The long shadows in the morning and evening take away from the usability of the image. Flying midday eliminates shadows but may degrade photos of concrete or light colored and reflective surfaces.

4. Don't skimp on checks. Fill in the gaps between your panels with red or similar colored painted dots. You can also use red plastic dinner plates with a nail in the center or print your own on mylar.

5. If you start generating contours, understand and use the 2023 ASPRS Positional Accuracy Standards and give due respect to the science of photogrammetry (I'm sure you will, but there are many sUAS Airmen who seem to think they know more than photogrammetrists or the FAA and can pick and choose which parts of the guidelines to follow).

I suspect that after a dozen or so flights you'll find a simple and quick workflow. I've stopped counting how many times measuring from my own ortho has saved an additional trip to the site.

Good luck

#4 · August 13, 2024, 5:39 pm

Thanks. That was one of the ones that kept showing up when searching for software. It certainly looks like it would accomplish the goal.

#5 · August 13, 2024, 5:48 pm

Thanks for the tips murphy. I'll take that last bit to heart - If I were to go into trying to develop contours, I would certainly be doing a great deal more reading and practice.

I've made a couple of orthophotos and laid them over some existing state published imagery. It looks pretty good so far. One item I like about the drone images is that the color is more vibrant (of course the state imagery is more functional as it was flown in winter).

I have not yet added ground control points into the mix. I'm planning to experiment with that a little bit soon and see how it may improve the RTK results from the drone. For what I am trying to accomplish, I've been well pleased with the results straight from the drone images without GCPs. But if there are improved results to be had by adding the GCPs, that will be incorporated.

#6 · August 14, 2024, 9:47 am

BTW, if you are in the Civil3d world, ACAD has a solution in ReCap for processing photogrammetry.

#7 · August 18, 2024, 1:15 pm

WebODM is the cheapest, easiest way to try. Aerotas, last I read, uses Global Mapper Pro for their feature extraction. Civil3d can be used if you already have it, though, it is not the best option. You can convert things to .RCS using the free version of Recap.

#8 · October 6, 2024, 10:22 pm

"Aerotas" does not use Global Mapper Pro. I am one of those engineers who work on your projects and have done a lot of them 🙂

For 2D Planimetric mapping you can use any CAD (Bentley MicroStation, AutoCad, Autodesk Civil 3D) or Agisoft Metashape, 3D Planimetric mapping is more complicated, you can work in Bentley MicroStation if you have experience.

The only thing I want to draw your attention to is that to form a correct TIN model you will have to work very hard on breaklines and check the height marks of each point, usually on DEM, DTM, DSM a lot of poor quality data, a lot of heavy vegetation (especially in summer), a lot of dead zones with errors (orthorectified)

Thx