I have organised an introductory day of events for lower sixth students this week (Friday 29 June), aimed at providing taster practical sessions on GIS, as well as trying to logically link together schools (and subjects), universities (and degree choices) and GIS professionals (and jobs). The day is based around two practical sessions run by my colleagues James O’Brien and Ken Field. This is then followed by a dicussion panel that I am chairing, and we are grateful to the following for giving their time to be at the event:
The day is finishing with a keynote presentation by Vanessa Lawrence, CEO and Director-General of the Ordnance Survey (and a Visiting Professor at Kingston). We are honoured to be able to welcome her to the university. More news on all the events of the day next week.
Scott Haefner had an interesting idea in combining a fisheye lens with kite aerial photography to create some 360 panoramas. Strangely Scott and I fly exactly the same cameras with a Nikon D70 and Coolpix 8400. The latter can take the Nikon fisheye lens with the aid of a converter. And it takes great pics. Scott takes a “vertical down” photo from the kite and a “vertical up” from the ground, “unstitches them” and then combines them together before putting them in to an interactive photo.
What interested me was the unstitching or unwrapping part. The lens has a 360 degree field in the vertical plane and 180 degree field of view in the horizontal plane. In essence everything in front of the camera is in the photo because it is an ultra-wide angle lens and the picture you get is a circular image (think of it as more akin to a “dome”) in the middle of a standard photo frame. OK, now point the camera straight up and take a photo. You get a picture of the sky with athe entire horizon wrapped around the outer edge of the circular picture. Now mark the centre of the photo and cut directly in to from the outside edge. Finally “unwrap” the photo, correcting for the lens distortions between the centre and edge, to produce a panoramic shot. This is essentially what the software is doing.
So what do we use to create the panoramas? Well the primary package is a command line opensource product called PanoTools. Numerous vendors have created GUI front ends (PTGui being notable), including a late beta open source product called Hugin. Whilst having a few rough edges it is really very good and works without problem for KAP purposes. PanoTools is designed for creating all sorts of panoramas, not just from fisheye lenses. In fact it models lens parameters. One important thing to note is that the lens model for the fisheye is not suitable (only 160 degrees FOV) and it is therefore necessary to get a patched DLL file (full 180 degrees FOV). A good guide (using PTGUI) can be found here. More on using Hugin in another blog.
In fact installing GRASS isn’t that difficult, but for those not having used Linux before it is, on the one hand, intuitive and much better than Windows, whilst on the other, if you don’t know how to do it, it won’t work! So in short (and thanks are due to Carlos for the pointers):
1. Go to “System” and “Synaptic package manager” 2. Go to Settings -> Repositories. 3. Enable all the repositories (universe, multiverse, restricted etc) 4. In the “Third-party software” tab, add a new repository using this line:
deb http://les-ejk.cz/ubuntu feisty multiverse
5. Reload the packages list and search for “grass”. Double-click onits name, the package manager will ask if you want to install also thedependencies (yes!). 6. Its worth installing Quantum GIS (aka QGIS), part of the OSGeo stable, as it’s a great data browser/editor with some analysis functionality. It canalso be used as a front-end to GRASS (install “qgis” and the “qgis-grass-plugin”).
I have played around with various Linux distros over the years (and remember installing BeOS at one point!), and whilst they look nice, when it comes down to doing “real” work I end up getting stuck. So much of my software is Windows based, particularly for a lot of bespoke applications. And of course, with my move to primarily using USB apps, all the places I visit also use Windows. To be fair its less of a problem now with Linux running software like OpenOffice, FireFox, Thunderbird, GIMP etc. That said, Linux is still popularly a server system. However there have been big strides to make Linux cuddly, friendly and generally just as user-friendly and easy to install as Windows or MacOS.
The distribution that has been really helping here is Ubuntu and its lighter weight siblings Xubuntu and Edubuntu. These really are easy to install and use with everything you need for a “typical” home office system. What has really shifted me in this direction is the long term open-source GIS, GRASS. This is available for Windows, however the Linux version is the primary product that is more stable and has better features. So this has shifted me in the direction of Linux, but how best to use it??
Well the obvious answer is to install it! This could be on an old system, on my main system in dual-boot mode or with a removable hard drive. All of which really involves too much fiddling about. Which brought me back to our old friend virtualisation that we have been using to run ArcIMS in the student labs. There are two main products, VMWare and Microsoft Virtual PC, which are vying for trade and both offer free, full functional, versions. I tried both, but in the end Virtual PC was faster and more stable so won out. Any modern PC should be able to run Ubuntu or Xubuntu pretty fast with no desperate problems, and being virtual you can just kill the system and start again if there are any problems. You need to go over to the Ubuntu website and download the latest ISO image which can be loaded in to Virtual PC as a CDROM drive, booted from to give a “Live CD” running Ubuntu and then used to install a full system. Xubuntu kept freezing on me, so I stayed with Ubuntu.
Unfortuantely the latest Linux kernel has a bug which renders the mouse unusable. A quick Google brought up a a fix which, whilst a pain, at least gets things working. It sets up a keyboard mouse, which allows you to install the system. You then drop back to a command prompt, load in a new patch and then reboot with it all working. In particular, when you get to a reboot you need to follow the commands below:
1. reboot in recovery mode by pressing ESC on the GRUB screen 2. wget http://librarian.launchpad.net/7583925/unsupported-patch-for-87262.sh 3. chmod +x unsupported-patch-for-87262.sh 4. ./unsupported-patch-for-87262.sh 5. reboot
You should now have a fully virtual Ubuntu system sitting in Virtual PC into which you can install GRASS (and it should be hooked up to the network through your PC). However thats a story for the next blog.
“Zoom to Selected” is a handy ArcMap tool. Once you select a feature, just click on this item in the Selection menu and it fills the data view with the items currently selected. Very useful for locating small features or running a query and then looking at where they are spatially located.
Except it doesn’t appear to work for raster layers. My son asked me a classic GIS question: “What is the highest point in Bedfordshire?” I could have just Googled it, but the GISer in me simply had to do this inside a GIS. A visit to Digimap allowed me to download a selection of OS Panorama raster DEM data. I also grabbed the county outline from UK Borders. This I used to clip the DEM (using Hawths Tools as Spatial Analyst seems to make such a pigs ear of it) to leave me elevations of Bedfordshire.
It was then simply a case of sorting the elevation column on the attribute table and selecting the highest value. A trip over to the “Zoom to Selected” and… it didn’t work. ESRI support suggests this is the case. I must have missed something blindingly obvious, but I can’t see a simple, obvious, way of quickly locating the highest point. In the end I zoomed in and scanned around the image until I found it, copied the coordinates and checked them on Streetmap. OK, so it wasn’t right bang on the (old) trig point on Dunstable Downs, but it was pretty close. Not bad when you think these are 50m pixels generated from contours.
I use the “Assignment” feature on BlackBoard extensively for students to submit material. Not only is there no “physical” copy to print, submit, collect, mark and return, but the student can submit from home and I can collect them from home. Winners all round (and because its digital it means that we can screen it for plagiarism, but thats another story).
This is great when it works. And an absolute pile of **** when it doesn’t. The last thing I want to end up doing is finding corrupt files in the middle of marking and then spend days chasing students for coursework. Unfortunately this is something that has happened all year and seems to be caused by large file uploads. Staggeringly, BlackBoard doesn’t employ file checksums to ensure that transfers are error free. My rudimentary understanding would suggest that it simply receives the stream of packets and stores this data. This isn’t too much of a problem for small files where the likelihood of errors is low. However it is a huge problem for large Word documents and ZIP archives. The latter are a real issue as when the archive becomes corrupt nothing can be accessed. For the classes I have been dealing with, error rates have been running at 10-20% which is simply not good enough. This really is a show stopper and something that urgently needs addressing.