So the data are free for anyone to look at. But I'm not aware that there's any way to download georeferenced imagery from Google Maps? I mean, they're making the imagery available, but it doesn't seem all that useful to me from a photogrammetry standpoint. You don't have nearly enough information to do a lot of kinds of analysis using just a color-balanced RGB image (that may have been through some lossy compression process?). It seems like your Ikonos data are still of superior quality and use to what can now be seen on Google Maps. So what's the problem?

It used to be artificially limited (they called it 'selected availability'). Today, US GPS has selected availability turned off, so civilian GPS users have access to the same data as the military. I don't believe there's any technical reason why they couldn't turn it back on, but GPS has proved to be so useful for civilians that it'd probably have to be a pretty serious situation that would prompt them to do it.

Stationary GPS is a little bit different. The receiver is planted in a location whose coordinates can be very carefully determined via more traditional survey methods. Combine this with some other technologies, and you can get very precise and accurate results. For example, one of the factors that degrades the accuracy of GPS is atmospheric effects. With a network of carefully surveyed stationary GPS units, we can correct for atmospheric effects by seeing how 'off' the various units are compared to normal, and to each other. There are other sources of error, but the point is that GPS error can be greatly reduced when you already know where you are.

Now, in this case, the 'stationary' GPS units are actually moving at a very slow rate. With the error corrections described above, once all the other errors are accounted for, what remains is error due to actual movement of the GPS. I can't see the full text of the paper, but probably what they have is a statistical model that says the GPS units are moving by a certain amount each year, and a confidence level, and all of that.

So, to your last point: if you want to improve the GPS accuracy of your lawn bot, you need only to install a stationary GPS receiver on your house, survey its location very carefully, and attach a transmitter to turn it into a 'GPS base station' that your robot's GPS will use as a local reference to improve its GPS fix. (You can buy a GPS base station from someone like Trimble; they're often used for construction and the like.)

In most cases, including this one, the prisoner is interested in using his own money to buy an Xbox, not tax dollars.

Er... actually, President Bush toured the country in a bus from the same company.

404 pages that don't return code 404 make me crazy. I worked on some software that did periodic harvesting of remote data, and there was one site that was always moving its files around, and they had a custom 404 page that returned 200 OK. So the software was never able to tell that the file it was looking for wasn't actually there, and our database for that site was always screwed up. They would contact us and complain about our links being wrong every now and then, but I could never get them to fix their 404 page. To be honest, I'm not sure I ever got them to understand what the problem was.

The point of the whole project was to automate the process. If we have to go and visually inspect every URL to make sure that the remote web server isn't lying to us about the validity of the link, then it kind of invalidates the whole thing.

Probably not from Microsoft, but the OpenKinect community has developed an API that works under many platforms:

http://openkinect.org/wiki/Main_Page

Okay, that explains it. The submitter was only willing to put in the smallest possible amount of time and effort to make his blog post not so transparently an interstitial page with no beneficial contribution to the topic.

I don't mind if people link to their own blog in the summary, but if they're going to, at least they should make their blog post *useful*.

Thanks.

Probably, not every single computer in the OP's work needs to work with a Kinect. For some applications, it would be enough to have one machine dedicated to that kind of application, and its use would be shared among the office.

The blog post must have been written in a hurry by someone whose native language is not English.

Here is the link to the actual Microsoft SDK:

http://research.microsoft.com/en-us/um/redmond/projects/kinectsdk/