Top Document: Satellite Imagery FAQ - 3/5 Previous Document: Is there a program to compute Assessment measures, including Kappa coe Next Document: I need to classify a mosaic of several images. How best to do it? See reader questions & answers on this topic! - Help others by sharing your knowledge How good are classification results in practice? The following detailed commentary was posted by Chris Hermansen (clh@tfic.bc.ca). Mike Joy posted a question regarding irregularities between two classifications, one derived from manual interpretation of large-scale aerial photography, the other from a supervised and enhanced spectral classification of Landsat TM imagery. I've read several of the responses, and I just thought it time to kick in my $0.02 worth, since I am quite familiar with both of the classifications with which Mike is working. First, Peter Bolton rattles off his experience in tropical forests and chastises Mike for discovering what should have been obvious. Well, Peter, the boreal forest is a much different beast than what you're used to in Malaysia (I can attest from firsthand experience in both cases). Classification from remotely sensed data is generally quite reliable in the boreal forest, especially given the vegetative nature of the TM-derived classification that is Mike's second dataset. Detecting predominantly deciduous from predominantly coniferous stands is (spectrally speaking) pretty straightforward. Problems arise in mixedwood stands, however, since the nature of the classification of proportion is not necessarily the same and in any case any aggregative techniques applied to the TM image prior to classification (eg smoothing) could significantly alter the proportional balance. Also, depending on the proportion of deciduous in a predominantly coniferous stand, and the spatial distribution of deciduous trees within that stand, the classifier may have difficulty detecting the differences between mixedwood and younger pure coniferous types. Furthermore, deciduous stands with coniferous understory are classified as deciduous in Mike's first dataset but may easily be interpreted as mixedwood stands in the TM image. Secondly, on the subject of incorporation of field data, Mike's second dataset has some ground truthing incorporated in the classification. Thirdly, on the subject of large numbers of classes in some people's TM-derived classifications, remember that in many cases these additional classes are derived by incorporating other datasets (field measurements, other digital map data, DEM information, etc). The people I've seen most test this envelope are the folks at Pacific Meridan Resources; their TM-derived datasets form only the first step of several. As Vincent Simonneaux points out, most people stop at the first step. So, in response to Mike's original questions: > 1) Is it reasonable to expect a TM-based classification to accurately > distinguish Coniferous and Deciduous forest? The area I am dealing > with is boreal mixedwood forest in northeren Alberta, Canada. I had > expected that the classification should at least be able to do this. On the face of it, yes. But! You must ensure that your definition of Coniferous and Deciduous forest is exactly the same in both cases (and the prevailing definitions in use in Alberta don't exactly help out in this case). > 2) Do people out there have similar experiences, i.e. the actual >classification > accuracy being very much lower than the reported results, or major > differences when comparing with different source of information? Of course, this is a possibility; the most unreliable classes may interfere in a nasty way between to datasets. You really need to ensure that you are sampling the same population in both cases; then you need to examine the distribution of errors among classes in both cases. In your first dataset, you don't really have error estimates with which to work. > I > understand that an air-photo-based forest inventory and a TM satellite >image > are measuring different things, and that I shouldnt expect perfect >agreement, > but I would have thought they could agree roughly on the overall area of > Coniferous or Deciduous forest. Ditto for two similar TM-based > + classifications. Once more, not necessarily. See the points above on coniferous understory in deciduous stands and the basic definitions of coniferous/deciduous split. There are, of course, really obvious errors that can occur, like using pre-leaf or post-leaf images when trying to locate deciduous stands... Sorry to go on at such length about this; I hope that my comments are of interest to some of you. User Contributions:Comment about this article, ask questions, or add new information about this topic:Top Document: Satellite Imagery FAQ - 3/5 Previous Document: Is there a program to compute Assessment measures, including Kappa coe Next Document: I need to classify a mosaic of several images. How best to do it? Part1 - Part2 - Part3 - Part4 - Part5 - Single Page [ Usenet FAQs | Web FAQs | Documents | RFC Index ] Send corrections/additions to the FAQ Maintainer: satfaq@pobox.com
Last Update March 27 2014 @ 02:12 PM
|
Russians who want to impress their social media followers can now rent huge bouquets of flowers for just long enough to snap an Instagram worthy photo, It's revealed.
Pop up services are advertising on social networks in planning for International Women's Day on 8 March, A public holiday in Russia when women are in the past feted with flowers and other gifts. They're offering 10 minutes with an enormous bouquet sufficient time to perfect the best angle and pose before the courier takes it back, The TJournal news blog reports. One account offers women a fleeting visit from 101 roses for 700 roubles ($12; 10).
Posting selfies with huge bouquets seemingly sent by a boyfriend or secret admirer has been a trend among Russian Instagram and VKontakte social network users for quite a while.
The 360 TV website contacted the master of one flower rental Instagram account, [url=https://charmdatescamreviews.wordpress.com/tag/hot-russian-women/]russian sexy girls[/url] Who insisted his service was real and had received many asks. He says the 10 minutes can be drawn out a little, But not consistently. "the most important thing to avoid is: 'I'm obtaining put some make up on and tidy up' and all that', according to him.
Another account that 360 TV contacted ended up being a joke set up to mock the new trend, And there was plenty of ridicule from Russians placing comments online. "the secret of a mysterious bunch of flowers from a stranger has been EXPOSED, Writes an individual on Twitter. "might be quintessence of modern values, Says an Instagram surfer, Who suggests similar service aimed at men: "A Rolex download, Two iPhones shared and a selfie taken in a Moscow office with a panoramic view.