- TIN based - This technique relies on indentifying the 3 nearest non-planar points to the point being interpolated and fitting a planar or higher order surface to these points. The grid-point is then calculated as a point on that surface.
- Inverse Distance Weighted - This technique uses a weighted linear combination of all or a local subset of the known unevenly sampled points. The weights are calculated using various techniques e.g. 1/distance , 1/(sqr(distance)), exp(-sqr(distance)). If there are known characteristics in the sampling e.g. there is denser sampling in one direction, the weights can be biased to reflect this.
- Thin Plate Splines - Similar to the IDW techniques splines can be fitted to the full or local subset of the unevenly sampled dataset. Splines produce smooth interpolation results, but can have large spikes in sections if there are inherent discontinuities in the underlying data or large gaps.
- Geostatistical Techniques - Foremost of this is the Kriging technique which makes assumptions about the parametric distribution from which the sampled points are drawn and estimates the parameters of this distribution using maximum likelihood estimators. The assumption of spatial dependence may not be valid for widely spaced samples and statistical tests should be employed before the interpolation is used.
Monday, August 31, 2009
I haven't been there - so I will just Interpolate
Though not as hazardous and short sighted as extrapolation, interpolation can be quite challenging and fraught with errors and pitfalls. Whenever data is collected on the ground it is not evenly sampled and smooth, there are system drifts as well as uneven motion of the sampling platform. In order relate this ground data to nicely gridded remote sensing data, from SAR or Optical platforms, interpolation is needed. There are lots of interpolation techniques, I will discuss some of the easier to explain ones here:
Sunday, August 30, 2009
Organizing Xtreme Programming - while riding a monocycle on a tightrope
Somehow I have got myself into organizing the IEEEXtreme 24 Hour Programming Challenge. I innocently sent an email to the chapter chair about it, now it has landed on my lap to think up some programming challenges. I will appreciate ideas.
Coding WorldWind will be great but they only support Mono and Java so it is definitely going to be for WorldWindJ. I have some interesting ideas I would like people to take a shot at. Other than that there are classic algorihmic excercises of sorting and searching, may be searching for a possum in Flickr public image libraries with the code implemented on Google App Engine or Amazon EC2 , just for the kicks.
Sunday, August 23, 2009
Wine trip Barossa August 2009
Friday, August 21, 2009
Inglourious Basterds - Let's Get Em
Very nice QT movie, albeit a fantasy unlike Valkyrie ( I can't stand Tom Cruise though). Brings out the monsters that are created during war. My favourite character in the film is Hans Landa for sure, he is comic smart and even to some extent caring for people under his command : Radio Op. He shows ratlike survival instincts that he is so proud of and a shrewd read of the tide of war.
Thursday, August 20, 2009
Staying coherent - come rain come sunshine
SAR is widely touted as an all-weather system, but weather can limit the applicability and vastly affect the interpretation of SAR data. Not only is tropospheric weather we are familiar with the cause of concern, at longer wavelengths ionospheric effects and Faraday rotation can become an issue, impacting polarimetry.
One of the impacts moisture in the atmosphere is attenuation and path length extension as the electromagnetic properties of the transmission medium change. This can impact interferometric applications.
In addition to instantenous effects on the wave propagation, weather be it wind, rain or may be a long spell of dry weather can have complex effects on the target backscatter. Grass may dry up and disappear from view. Canola crops and trees may wave and become motion blurred ( saw some marvellous examples of that at IGARSS). Rain and hail may saturate the ground making things suddenly darker/brighter. In such cases direct comparison even accurate coregistration becomes difficult. Theoretical modelling and experiments put the wind based decorrelation of trees to a few tens of milliseconds, putting limits on integration time achievable.
All this makes repeat pass interferometry very fickle just like the weather. Coherence can only be achieved in very limited areas.
One of the impacts moisture in the atmosphere is attenuation and path length extension as the electromagnetic properties of the transmission medium change. This can impact interferometric applications.
In addition to instantenous effects on the wave propagation, weather be it wind, rain or may be a long spell of dry weather can have complex effects on the target backscatter. Grass may dry up and disappear from view. Canola crops and trees may wave and become motion blurred ( saw some marvellous examples of that at IGARSS). Rain and hail may saturate the ground making things suddenly darker/brighter. In such cases direct comparison even accurate coregistration becomes difficult. Theoretical modelling and experiments put the wind based decorrelation of trees to a few tens of milliseconds, putting limits on integration time achievable.
All this makes repeat pass interferometry very fickle just like the weather. Coherence can only be achieved in very limited areas.
Wednesday, August 19, 2009
The fallacy of classification
Classic remote sensing classification type problems try to lump ground features, lets say vegetation cover, according to artificial man made boundaries. These boundaries never model reality and hence the premises used to perform classification are false, this makes supervised classification unpopular. The other approach is unsupervised classification, where clustering techniques are used to identify naturally occuring variations and groups in the data; classes are built from clusters so identified. After this again comes the human intervention and learned or in some cases guesswork assignment of vegetation type to each class.
In between the 2 extremes are semi-supervised approaches, how strong the human intervention is determines the quality of the output from the classification problem. All this does not address the problem that classification itself is a fallacy, an ill posed problem and groups are not well modelled for algorithms to yield desired results.
Consider the data fusion scenario shown below with multi-temporal, multi-polarization, multi-wavelength SAR and Optical images. First and boring housekeeping is to register all the images together, which can be done after fair bit of fiddling, picking ground control points by eye (automatic registration does not work well across L-band SAR and Optical) and choosing a suitable non-linear transform to map from slant range to ground range ( why can't the PALSAR polynomial be easier to access and more accurate ?).
Consider the hypothetical problem of classifying river, vineyard, centre pivot irrgation areas and scrub. The vineyard alone as a class is not well defined, there are leafless vines which are virtually invisible in optical but visible in x-band SAR. The river has bright pixels in SAR which are tree trunks sticking out of the water forming dipoles. Pure pixel based statistics will fail at classification here. What can be achieved is the appreciation of the variablity and the detail visible in each cover type at different wavelengths.
In between the 2 extremes are semi-supervised approaches, how strong the human intervention is determines the quality of the output from the classification problem. All this does not address the problem that classification itself is a fallacy, an ill posed problem and groups are not well modelled for algorithms to yield desired results.
Consider the data fusion scenario shown below with multi-temporal, multi-polarization, multi-wavelength SAR and Optical images. First and boring housekeeping is to register all the images together, which can be done after fair bit of fiddling, picking ground control points by eye (automatic registration does not work well across L-band SAR and Optical) and choosing a suitable non-linear transform to map from slant range to ground range ( why can't the PALSAR polynomial be easier to access and more accurate ?).
Consider the hypothetical problem of classifying river, vineyard, centre pivot irrgation areas and scrub. The vineyard alone as a class is not well defined, there are leafless vines which are virtually invisible in optical but visible in x-band SAR. The river has bright pixels in SAR which are tree trunks sticking out of the water forming dipoles. Pure pixel based statistics will fail at classification here. What can be achieved is the appreciation of the variablity and the detail visible in each cover type at different wavelengths.
You can work it out by Fractions or by simple Rule of Three,
But the way of Tweedle-dum is not the way of Tweedle-dee.
You can twist it, you can turn it, you can plait it till you drop,
But the way of Pilly Winky's not the way of Winkie Pop!
- Rudyard Kipling (Jungle Book)
Saturday, August 15, 2009
Oh pretty flowers !!
Finally back from a couple of days of running around in wheat, beans and other fields. Of course dodging suspicious country folks and whistling bullets ... just joking. I just got stopped once by a guy in a red car, he only took my card and said he will check with my office whether I am real. He also did not seem worried about some tiny samples. Saturday was a great day for the field, sunny and not so cold and windy. Covered a fair bit of ground and grabbed as many points as I could.
Then managed to make it to Walaroo in time for sunset and dinner, which was a bad idea since someone took this opportunity to steal my car park in the hotel and I had to park in the street. Sunday morning turned out to be windy and rainy and super grey. So I managed to do only a couple of points. On my way back near Balaklava (that name still cracks me up), I found a beautiful canola field worth exploring ... er if you are crazy like me.
Monday, August 3, 2009
Meeting Maurizio
I had an eventful arrival in Jo'Burg. The cab driver had no clue where my hostel was, even though it's a stone's throw from the airport. I had to whip out the iPhone and guide him. Come on Google make routing work in South Africa, people will so use it during the World Cups-2010, especially since cabs don't have a GPS. Anyway I got distracted and left my laptop in the cab. More Googling for the airport number followed .... eventually I got it BACK !!! Apparently a 1st in Jo'Burg.
Then I hopped on whirlwind tour of Johannesburg and Soweto, which ended at the Hector Pieterson museum. Then I nearly got left behind by the tour bus while shopping for souvenirs.
Finally the tour dropped me off at Maurizio's place and we spent a fair few hours discussing life, economics and tile serving. Thanks a lot for your hospitality Maurizio, I had a great time.
Then I hopped on whirlwind tour of Johannesburg and Soweto, which ended at the Hector Pieterson museum. Then I nearly got left behind by the tour bus while shopping for souvenirs.
Finally the tour dropped me off at Maurizio's place and we spent a fair few hours discussing life, economics and tile serving. Thanks a lot for your hospitality Maurizio, I had a great time.
Subscribe to:
Posts (Atom)