Hello Fellow Yuneec Pilot!
Join our free Yuneec community and remove this annoying banner!
Sign up

Basic question on how mapping works

Joined
Dec 28, 2016
Messages
106
Reaction score
28
Age
65
What data is generated by flying a mapping mission? Anything special or just a collection of geotagged still images? Apologies for my ignorance on this, how is what is generated imported to something a professional land surveyor could use?
 
Joined
Jun 9, 2016
Messages
1,478
Reaction score
1,293
Age
49
Location
UK
It's the same as most other survey apps (Mission Planner, Pix4D etc. etc.). The Mapping mission consists of an area you want to survey and your settings for what resolution of images/how high you want to fly and how much overlap between images. That's the "Input" on the ST-16S. The "Output" is indeed a set of geotagged images which can then be processed by any of the standard survey tools - Agisoft, Pix4D etc.

The processing takes a while - an hour at the least (and overnight runs aren't uncommon), and if you're using a desktop app, you need a decent PC to run it. When the images have been processed you can end up with:

a) An orthomosaic - that is a single, stitched photo that is stitched and warped so that all vertical objects remain vertical. It can be overlaid on Google Maps or other survey data. For instance: Crucible The original image for this survey covers about 12 acres and is around 283 megapixels.

b) A point cloud - that is a few million 3D points that make up the surface of the area you surveyed. This can allow you to measure obects and volumes within the surveyed area. It's not easy to share online, but an example of an online viewer with some sample datasets is here: plas.io

c) A 3D textured surface - that is a 3D model using the point cloud and photos combined to give you a physical representation of the area being surveyed. You can view this in a number of different packages, including various online tools. As an example: Crucible Woods by tunauav - 3D model (Note that this is reduced resolution over the normal output - the generated file is a few hundred megabytes).

There are some other options you can play with - such as false colour and extended spectrum photography (used in agricultural surveys) and thermal scans.

For accurate processing you may need to generate a ground control point (GCP) file - which is used to precisely locate the geotagged images so that they produce more accurate maps. Without GCPs the surveys can be a a few inches off, which is less of a problem for some survey types than others. With ground control points, accuracy can be brought down to an inch or two - but getting below that threshold is difficult and expensive. Civil engineers will be disappointed to find you cannot get millimetre accurate models through aerial surveys. Creating a GCP file usually needs an accurate GPS locator, which is used to spot measure a number of reference points (usually specially marked mats) that allow points in the captured photographs to be precisely located.
 

arruntus

Premium Pilot
Joined
Aug 19, 2017
Messages
1,662
Reaction score
579
Very good explanation :)

With respect to the precision of the models it is possible to obtain millimeter precision, a GSD of less than 1cm can be achieved. I have processed a project, I did not make the flight myself, where the flight height was very low and I used a camera with high resolution and a sensor of 1 inch (now I do not remember the specific data and i erased it because it weighed so much). Was a very heavy and expensive aircraft, just like the used camera. Now with the H520 that incorporates a good quality camera I think we can get it too, hopefully.

It is very good to try to achieve very high resolutions but as a counterpart it is necessary to lower the flight altitude a lot which means that the footprint of the images is smaller and therefore the amount of images that must be taken to cover a terrain is multiplied exponentially and we pass to processing time of more than 24 hours in a row. Normally a GSD of between 2 and 3 cm is usually enough in the vast majority of cases, at least in what I know.
 

arruntus

Premium Pilot
Joined
Aug 19, 2017
Messages
1,662
Reaction score
579
Continuing on the same subject of precision, these are data obtained with the H520 and E90.

A 60-70 overlap (this is not certain because the data is not the same as when he configure the mission).
Area covered: 0.0203 km2
Flight altitude: 36 m
Number of images: 86
GSD: 8.93 mm/pix

We can see it in the video in the 9:56 minutes. (I don't know how to link to play it in a given second)

 
Joined
Dec 28, 2016
Messages
106
Reaction score
28
Age
65
It's the same as most other survey apps (Mission Planner, Pix4D etc. etc.). The Mapping mission consists of an area you want to survey and your settings for what resolution of images/how high you want to fly and how much overlap between images. That's the "Input" on the ST-16S. The "Output" is indeed a set of geotagged images which can then be processed by any of the standard survey tools - Agisoft, Pix4D etc.

The processing takes a while - an hour at the least (and overnight runs aren't uncommon), and if you're using a desktop app, you need a decent PC to run it. When the images have been processed you can end up with:

a) An orthomosaic - that is a single, stitched photo that is stitched and warped so that all vertical objects remain vertical. It can be overlaid on Google Maps or other survey data. For instance: Crucible The original image for this survey covers about 12 acres and is around 283 megapixels.

b) A point cloud - that is a few million 3D points that make up the surface of the area you surveyed. This can allow you to measure obects and volumes within the surveyed area. It's not easy to share online, but an example of an online viewer with some sample datasets is here: plas.io

c) A 3D textured surface - that is a 3D model using the point cloud and photos combined to give you a physical representation of the area being surveyed. You can view this in a number of different packages, including various online tools. As an example: Crucible Woods by tunauav - 3D model (Note that this is reduced resolution over the normal output - the generated file is a few hundred megabytes).

There are some other options you can play with - such as false colour and extended spectrum photography (used in agricultural surveys) and thermal scans.

For accurate processing you may need to generate a ground control point (GCP) file - which is used to precisely locate the geotagged images so that they produce more accurate maps. Without GCPs the surveys can be a a few inches off, which is less of a problem for some survey types than others. With ground control points, accuracy can be brought down to an inch or two - but getting below that threshold is difficult and expensive. Civil engineers will be disappointed to find you cannot get millimetre accurate models through aerial surveys. Creating a GCP file usually needs an accurate GPS locator, which is used to spot measure a number of reference points (usually specially marked mats) that allow points in the captured photographs to be precisely located.
Thank you for the very detailed and well thought out reply. Just what I needed to know! - Rick
 
Joined
Jun 9, 2016
Messages
1,478
Reaction score
1,293
Age
49
Location
UK
Very good explanation :)

With respect to the precision of the models it is possible to obtain millimeter precision, a GSD of less than 1cm can be achieved. .
It's worth pointing out there is a difference between resolution (the physical distance one pixel in your orthomosaic covers) and accuracy. Flying lower and using a higher resolution camera increases the detail - and these days you can easily go to 1/2" or 1cm per pixel, which is enough to spot small details in your map.

Accuracy comes from a combination of GPS sampling, lens distortion and stitching artefacts - and can be quite variable. The result is that a straight wall of 100m might measure as 100.5m on the orthomosaic, or appear curved or bent. Though the detail may be high enough to see the mortar between bricks, the accuracy may be far lower. It's this that causes problems for civil engineers, and why you might choose to use GCPs to improve things.
 
  • Like
Reactions: arruntus
Joined
Dec 28, 2016
Messages
106
Reaction score
28
Age
65
It's worth pointing out there is a difference between resolution (the physical distance one pixel in your orthomosaic covers) and accuracy. Flying lower and using a higher resolution camera increases the detail - and these days you can easily go to 1/2" or 1cm per pixel, which is enough to spot small details in your map.

Accuracy comes from a combination of GPS sampling, lens distortion and stitching artefacts - and can be quite variable. The result is that a straight wall of 100m might measure as 100.5m on the orthomosaic, or appear curved or bent. Though the detail may be high enough to see the mortar between bricks, the accuracy may be far lower. It's this that causes problems for civil engineers, and why you might choose to use GCPs to improve things.
More good advice, thank you. - Rick
 
Joined
Oct 30, 2017
Messages
16
Reaction score
1
Age
47
Accuracy comes from a combination of GPS sampling, lens distortion and stitching artefacts - and can be quite variable. The result is that a straight wall of 100m might measure as 100.5m on the orthomosaic, or appear curved or bent.
You also have relative accuracy which is relevant to the local scene and how the calculation interact on a scene level (volumetric, distance etc) and absolute accuracy which the geospatial accuracy against the actual known position on the planet - we quite often get relative to a few cm by measuring a few ground control points, and a ground resolution of .7cm/pix and absolute x,y accuracy of 3-4cm and z accuracy of 6-10cm.

We can get mm accuracy but that means leaving gnss data collectors on for over 24 hours per ground control point and PPK post processing.
 
Joined
Jun 7, 2016
Messages
781
Reaction score
224
Age
61
Location
Peak District, UK
A question from myself if I may.
I need to be able to produce accurate topographical contour drawings as in just the contours in black on white. If I use GCP's for accuracy am I correct that using Pix4D or Agisoft it is possible to produce such a drawing?
 
Joined
Oct 30, 2017
Messages
16
Reaction score
1
Age
47
@bluelight.support , do you mean you post processing the Ground control Points, or are you post processing the geotags on the images somehow?
I will use 4-5 ground control points in various areas of the mapping scene, run my mission with images being geo-tagged with the aircraft gps for 5m+ accuracy then post process. So ill get my ground control point data and post process to gm cm accuracy on the ground control points. Ill then import all the images to Photoscan Pro and basic process, then add the ground control points into the data, and manually optically line them up over 3-4 images, that will then give accuracy to 1cm+/- in xy, and around 5cm z
 
Joined
Oct 30, 2017
Messages
16
Reaction score
1
Age
47
A question from myself if I may.
I need to be able to produce accurate topographical contour drawings as in just the contours in black on white. If I use GCP's for accuracy am I correct that using Pix4D or Agisoft it is possible to produce such a drawing?
Yes - if you ensure that your ground control points are geo-accurate and match to your data, then when you create your contours they will all match as part of the data set. I only use Agisoft as dont like the workflow in Pix4D and also had support issues with the software.
 
Joined
Jun 7, 2016
Messages
781
Reaction score
224
Age
61
Location
Peak District, UK
Yes - if you ensure that your ground control points are geo-accurate and match to your data, then when you create your contours they will all match as part of the data set. I only use Agisoft as dont like the workflow in Pix4D and also had support issues with the software.
Thanks for the info @bluelight.support I am investigating its accuracy and use for our company and I will need to do some real world trials soon and with a bit of luck will get a nice bit of drone work out of it.
I assume you are using the H520 for this, E90? Would you consider E50 a useful option for greater definition or would it make too many images?
Why are you stitching in photoscan I thought Agisoft could handle that as well?
What are you using to set your GCP's?

Sorry about all the questions but its a new area for me.
 
Joined
Oct 30, 2017
Messages
16
Reaction score
1
Age
47
I do all the processing in Photoscan so do the initial processing, add GCP for cm accuracy and go from there - the workflow allows for point clouds, 3d models, orthophotos, contours, measurement volume, etc - no im still using phantom 4 pro as its mechanical shutter. Went from Inspire 1 Pro which was brilliant, to the Inspire 2 which is rubbish, and Phantom, still waiting for the yuneec to be a little more settled and functional before taking the plunge, I like the hex, but would also prefer 2 battery for backup.
 
Joined
Feb 27, 2018
Messages
171
Reaction score
37
Location
Brisbane, Australia
Thanks for the info @bluelight.support I am investigating its accuracy and use for our company
If you are after accuracy you may want to consider something like this
Reach RS+ — Emlid

It is almost $800 but a) it will give you a centimeter accuracy b) you only need one station.

With this RTK you can have just simple DIY tiles as GCPs and you can have as many as you want at almost no cost. However you will need to record coordinates of each of GCPs centers and then correct each tile position in your mapping software.

If you scroll down a bit on the page on the link above there is a picture of how it is done. So you will be doing the same at each GCP and recording each GCP coordinates.

Another option is just get an RTK module (Reach M+ — Emlid) from them and build a station like this yourself, Their RTK module is less than $300 and I believe you should still be able to use their app to connect to it and get all data from it.
 
Joined
Jun 7, 2016
Messages
781
Reaction score
224
Age
61
Location
Peak District, UK
I do all the processing in Photoscan so do the initial processing, add GCP for cm accuracy and go from there - the workflow allows for point clouds, 3d models, orthophotos, contours, measurement volume, etc - no im still using phantom 4 pro as its mechanical shutter. Went from Inspire 1 Pro which was brilliant, to the Inspire 2 which is rubbish, and Phantom, still waiting for the yuneec to be a little more settled and functional before taking the plunge, I like the hex, but would also prefer 2 battery for backup.
Thanks interesting, I was asking in another thread how mapping copes with uneven terrain like a hillside, if the mission is all photographed at the same altitude then all images will be at different distances from the ground, so how does this all stitch together accurately, or can the phantom maintain a set distance above ground level?
 
Joined
Jun 7, 2016
Messages
781
Reaction score
224
Age
61
Location
Peak District, UK
If you are after accuracy you may want to consider something like this
Reach RS+ — Emlid

It is almost $800 but a) it will give you a centimeter accuracy b) you only need one station.

With this RTK you can have just simple DIY tiles as GCPs and you can have as many as you want at almost no cost. However you will need to record coordinates of each of GCPs centers and then correct each tile position in your mapping software.

If you scroll down a bit on the page on the link above there is a picture of how it is done. So you will be doing the same at each GCP and recording each GCP coordinates.

Another option is just get an RTK module (Reach M+ — Emlid) from them and build a station like this yourself, Their RTK module is less than $300 and I believe you should still be able to use their app to connect to it and get all data from it.
Thanks, I will look closer at this when I get home from work
 
Joined
Jun 9, 2016
Messages
1,478
Reaction score
1,293
Age
49
Location
UK
Thanks interesting, I was asking in another thread how mapping copes with uneven terrain like a hillside, if the mission is all photographed at the same altitude then all images will be at different distances from the ground, so how does this all stitch together accurately, or can the phantom maintain a set distance above ground level?
I answered on the other thread, but I'll respond here as well - you want the images to be at different distances from the ground so that the stitching software can accurately calculate the actual terrain being measured. A proper mapping application is not simply joining photos together to make a large photo - it uses the different speeds that things move past the camera (parallax effect) to calculate the actual height of the objects being photographed. That generates a detailed surface model that is used to 'warp' the photos so that vertical things appear vertical in all parts of the image. In ideal circumstances this should eliminate all the effects of taking photos at close range with a wide angle lens, and produce a final image that is as though it were taken from infinitely far away (no distortion).
 
  • Like
Reactions: 10-8 and OnTheRopes

New Threads

Forum statistics

Threads
16,477
Messages
192,326
Members
19,129
Latest member
fabricedu41100