Sunday 12 April 2015

Community Mapping Project

This week the weather has finally decided to show up and flying season is now underway. One area of drone flying which many feel is going to be popular this year is Drone Mapping and I have been busy with my own project this week; making aerial footage for a community woodland project and attempting to record some data on the progression of this woodland and it's new path.

The area covered by the drone taking its photographs is roughly 36 acres in area. The landscape is quite varied with steep hillside and tall woodland making the planning of the drone missions more challenging than flat areas.

Here is the area as shown in Google Maps.

Sadly, Google map data is around the year 2003 in Summer. A lot has changed in this area since then.

Here is my data collected from the drone in 2015 in early Spring.

There are one or two imperfections but I am happy as a first attempt. I feel it comfortably illustrates the community work that has been done in the area - Tree planting & Path creating.

Before flying this area, I walked to familiarise myself and understand the terrain more clearly. I decided that I could more comfortably split the area into three and make three flights, then connect the collected images when all is completed.

Here is a sample of flight #1.

And once I had collected enough strips in flight #1 I could merge them in an image editor to make an orthographic image of area #1:

This was repeated over the following three days for the remaining two flight areas. The weather was very consistent. The flights were repeated at similar times in the day.

Here is area #1 combined with area #2. Below is the Google maps image of the same area.

Eventually I had a completed map.

I was very amazed by the outcome; I was sure that there should be more difficulties and problems and I find myself being too harsh on myself trying to find the flaws and errors. Being English, my brain is very well tuned to expect hard challenge & denigration but this seems to have worked.

In total there were 3191 images collected and processed totalling 1070MB or 1.07GB in size.

Here is a droidplanner screenshot of the mission #1 flight route:

I guestimated the width of the 'flight lines' to make the images overlap enough to be merged cleanly. I guessed because I have not found a preset for my camera in Mission Planner Grid v2.

People are asking if I will use 3D to make the map animated, however I did not have the possibility to add GPS coordinates to the images at this time, nor do I have a workstation computer to process such data, nor do I have the expensive 3D software.

Perhaps in the future some people will like what I can do, perhaps even ask me to help them for their own projects, and then perhaps I will be able to afford to perform the 3D mapping from the revenues from these project people. But I am a sceptic and as always I feel that obtaining money and funding is more and more impossible, and these type of things will will be undertaken by enthusiastic volunteers, and the financially privileged (I am risking becoming political)...

One interesting future possible project for me, could be the mapping of the close-by huge cemetery. Perhaps the local council might fund me to map this area and integrate Plotbox allowing some very creative uses for cemetery management and plot logging. Who knows?

Anyway, here are some useful links if you wish to understand & undertake mapping

I suggest also having some skill in image editing applications such as Photoshop or GIMP, and some video editing applications such as Sony Vegas, Adobe Premiere, Microsoft Movie Maker, and Quicktime.

It helps also if you know how to make waypoint missions with your drone, and also have and know how to use a gimbal with camera

Some people also ask me if I have georeferenced & orthorectified the images, besides struggling to pronounce these two words, I have not deemed it necessary to georef my images for 2D mapping, orthoreerfherfungjening of the images; I just flew at a consistent altitude, and defisheyed the images. 

If you would like to view the three flight missions and evaluate them you can interactively view a playback movie by clicking these links:
Movie Flight #1  ---and--- Droneshare #1
Movie Flight #2  ---and--- Droneshare #2
Movie Flight #3  ---and--- Droneshare #3

You may also take use of the complete set of images that I used for this project should you wish to practice map processing for learning and familiarisation. I feel there are two clear stages in drone mapping: a) flight & planning, b) image processing & presentation.

The link for the whole folder of images is available here. You should select 'add to drive' to send zip & download the folder. It might take a while (It's over 1GB).

3D, IR, NDVI & MultiSpectrometry can all be explored in the future. Here's a picture about those:
Here's a video let's dance:

I experimented with 3D mapping today and it is quite an achievement for myself, I'm quite proud, but exhausted learning too.. I learned to georeference photographs and use some cool mapping software. The main issue I have is that my computer is NOT IN ANY WAY purposeful for such heavy tasks - The software didn't like my computer at all, and I don't think I will be able to map larger areas until I get access to a better computer. Anyway,

Here's a video

The camera I used is a Canon SX220 and it does not have GPS. I had to find another way to write the coordinates to the camera photos. Here is a photo from the camera to show the sample quality:

Here is how it looks as as an overlay on Google Earth:


fferro said...

Hi DalyBulge,
Good work!
I think I can offer a good deal for both of us.
I have knowledge in photogrammetry and 3D, as I work in forensic animation long ago (many years).
I'm moving into the subject of aerial photography recently with a couple of UAVs poor quality. Taking my first steps with them, nothing about planned missions, manual flights only.
I would like to offer my help with lab work for your images. With Photomodeler and PhotoScan, Global Mapper, Photoshop, etc. as well as what I do best, working with 3D Studio MAX to make quality 3D landscape models and animations over them.
In return I ask you keep me informed of your progress and help me to solve my problems with flight planning and recording images in the field as they arise.
Who knows. Maybe after passing time we could do some work together complementing our expertice.
Fernando Ferro

TheDalyBulge said...

Hi, this sounds like an interesting proposal. Sure, why not.
If you need help with flight planning it is always best to learn, but I can help to solve some of your difficulties.
I'd be happy to see if you can make a model with my test image set:

I'd like to see how it looks using high settings.
Eventually, I'm hoping to georeference & map again the whole area for 3D.


fferro said...

I'll do my homework.
I see some pictures out of focus. Did you set to infinity when shooting?

TheDalyBulge said...

yeah, I think the out of focus images happened during the waypoint change - the drone rotated.

Mészáros János said...

Hi Ben,

Sorry to not write as immediately as I proposed before on
Your original images were processed using the highest settings in PhotoScan 1.1.5 and you can download the .psz file from here:
Unfortunately, there are gaps between your images, overlapping parts are too narrow.

I'm going to process your video frames too but I have to search a time slot in the schedule of our cluster.

DalyBulge said...

That is an amazing amount of detail. I could never render such number of points & polygons.

The 16 image sample was my first experiment using a canon CHDK camera and using APM to trigger and georef the images. I feel the sample test was successful, so I will be attempting the total original map area using this method, but when the weather is good enough to do so.

It is very clear to me now, the difference in detail that maximum render settings brings to mapping, so thank you for helping me to have a sample to observe. My computer appears to have no problem displaying the result or high render, but it cannot perform the actual render.

Do you think the video frames will render? I am not so sure (as they have no georef).

Mészáros János said...

Yes, they will. I did very similar before and it performs very well.
The main advantage of Photoscan is that it doesn't require georef images. You can process images, create point clouds (sparse and dense) and 3d model and after that georeference them using ground control points, measured with geodetic survey tools e.g. GPS or total stations.