Is it possible to attain 1:1000 level precision from high resolution satellite stereo imagery?

Written by
Carlos Naranjo
Is it possible to attain 1:1000 level precision from high resolution satellite stereo imagery?

Remote Sensing Satellites are great because of its availability, cost-effectiveness, and easy to program tasks. On the other hand, in most countries, you have to go through a lengthy process to be able to take an aerial photograph (digital or analog) if you use an aircraft. This is one of the many advantages to use satellites instead of aircraft and also It is more practical from the project standpoint. But sometimes you need high-resolution imagery to perform engineering or cadastral grade work. Would it be neat to be able to do it with satellite imagery instead?

Remote Sensing Satellites are great because of its availability, cost-effectiveness, and easy to program tasks. On the other hand, in most countries, you have to go through a lengthy process to be able to take an aerial photograph (digital or analog) if you use an aircraft. This is one of the many advantages to use satellites instead of aircraft and also It is more practical from the project standpoint. But sometimes you need high-resolution imagery to perform engineering or cadastral grade work. Would it be neat to be able to do it with satellite imagery instead?

Aerial imagery has been for more than 150 years the primary source for cartography and other products. First in balloons, then in aircraft and now in satellites it is used to see and measure the earth from above. I did know little about it 30 years ago when i first encounter myself in a technical environment in which I had to digitize maps and then some aerial photographs that came from 23 cm wide film.Those days were wonderful in the sense that although I liked geography since I was a kid, I had no clue how all this was made. And that means that you have to learn and find out things for yourself without being contaminated by decades of engineering practices in the field of Geodesy and Photogrammetry. Don’t take this in the wrong way: without those two it’ll be impossible to be in the position we are today.

So back in the 90’s everything was on paper, photograph or film and it had to be digitized in order to derive digital maps, elevation models, and many other things in a more productive way. Computers were getting better and better but still lacked the sufficient power to perform production grade jobs without a huge investment in hardware. We were a small firm with two engineers and no funding so we had to either limit our scope to applications that does not need a great deal of costly hardware or we had to solve the problem in a way that would be both productive and cheap. We did the second, and this was done in that period in which everything were digitized so our first products were digitized maps, then we tried to update those maps using digitized aerial photograph and, fortunately,that happened in a place where the relief is not flat (My town, Caracas) so every effort I made to fit aerial photographs to maps failed until I realized that the only way to properly fit the digitized photograph is to reconstruct what happened when you take the photograph. So I wrote our first orthophoto program which takes the photo, the camera parameters (Thank god I was told by a Geodetic engineer to do that) then we keep growing that software until it became our workhorse. All that we did and do regarding cartography was made using tools developed in house. And we did that starting in the 90’s when, upto my knowledge, there were no photogrammetry technology developed in latin america. I don’t want to brag. It is just that we had to made thing from nothing, at least software wise. So we developed a whole suite of digital soft copy technology based on digital aerial stereo pairs. But more and more often it was hard to execute a project involving aerial photographs because of the lack of local availability of the aircraft with the appropriate permits and the aerial camera. Even if we had one the permits back then took several weeks and you could miss the imagery window (In a country without seasons it is hard to get an opportunity to take good aerial pictures)

Then the satellite stereo pairs became available and that changed the whole administrative picture overnight. If the imagery does not exist already which is up to date enough for the job at hand (normally the case back then) then you easily contact the satellite provider and program a satellite pass (tasking) so it will cover the are you need. It worked like a charm except that the satellite was too far away and the best you can offer is 1:5000 scale products.

But technology and research solve the problem eventually. Around 2014 new high resolution satellites were available such as the GEOEYE-1 and Worldview-2 and those satellites had better resolutions and much better accuracy on its RPC parameters (Rational Polynomial Coefficients). The only thing was that back in 2014 the best resolution a satellite company can give you was 0.5m not because the sensor has that much resolution but because US government won’t allow it for civilian applications. As of now that restriction is gone and you can get 30 cm imagery from a worldview-3 Satellite.

1:1000 Scale orthoimage derived from Geoeye-1 stereo pair

So the problem for us was how to use that wonderful sensor which will allow us to overcome the administrative burden which most of the time render our chances to compete useless and at the same time being able to match the technical specs of the job we were asked to do?

The answer came from a paper I read around 2009 speaking of the potential precision that can be reached using high resolution satellite stereo pairs such as Geoeye-1. That paper and others point into the right direction in the sense that it was worth the effort to develop the software and take the risk offering the service to a client. So we did that. In fact we offer –just to be sure – 1:2500 scale precision maps and the results were much much better. We got 1:1000 second order mapping precision using the Geoeye-1 stereo pair. In the following table you’ll see the kind of result we got.

Table 1. Absolute Orientation Adjustment results and checkpoint comparison

Well, you can say, Gee it’s awesome, well yes it is but the catch for us is that it completely eliminate the necessity of an aerial photography for all the scale range we offer. The same material can be used to derive regional, planning and high res, high precision material.  Who ever is on engineering planning for any energy or infrastructure project will stop here and read this again.

Also the casual reader can say: You might have succeed here but this is no reproducible.

Well, we have been using this methodology consistently since then up to the point that we have never use aerial photograph again if involved new imagery!

Please keep in mind that this methodology has its limitations. Obviously satellite Imagery has not the same quality at these scales as a Vexcel Digital Aerial Imagery for example (which by the way I think it is awesome) Also, sometimes there is no point in asking to task a whole satellite stereo pair - with a minimum of 100 sq km - if you are going to survey 50 hectares. But this is a practical solution and a very cost effective one if you want to survey in any part of the world and if you have more than one scale objective. Which is very common in engineering projects.

And the sensor have been getting better and better so this is just going to be easier to do. We did that when the raw imagery supplied to us was 50 cm per pixel  and when the super-resolution  algorithms were not that good. Now we have access to 30 cm per pixel imagery and far more better super resolution algorithms based on Deep Learning (AI). So the results with better sensors and better Hallucination (Super-resolution) algorithms is even better. That means reaching 1:1000 cadastral grade 1 mapping and topographic mapping with satellite stereo pairs in a consistent way.

Log in to view the Premium content in this report.
Connect with Author
Domains
GeoMatics
Satellite Imagery
GIS
Remote Sensing
Geodesy
Geospatial
Geography
Global
Venezuela