I was playing around on Google maps, and I clicked on ‘Street View’.
OK, one picture of a street, or even a house, would be easily explainable. But, what I saw was a 360 degree panorama of the street. Also, I could lift the pic up to see some of the telephone wires, well overhead!
I did note that my house had a vehicle in front of it that I got rid of at least 6 months ago, so I know that the image is old, but, I can’t figure out about the multidimensional image. How do they do it?
They take a bunch of images, with multiple cameras pointed in a number of directions (including upward), and stitch them together, more or less.
edit: In case you don’t know, they drive a car down the streets that looks like this and I’m assuming the camera takes pictures at timed intervals which get pasted together later on.
In addition to those special cars, I think Google is also using bicycles (or tricycles?) as well as other modes of transport to get all the areas possible (like alleys, parks, etc). That software that stitches the images together is pretty good at it, eh?
Google has been caught out in Germany. It appears not only has Google been taking images during street views operations, but also recording wireless network details. Google also claims they’ve done this in other places and have no plans to publish the WLAN details nor ties images to WLAN information.
Well, if there are no plans, why are they collecting the information in the first place? Are they doing it on their own or collecting the data for some other entity?
The article said they were going to target advertising with it.
So if the router you’re using is named “RON PAUL BLOWS ASS AND I <3 ABORTIONS” (actual one I’ve seen), and the Google camera cars have noted the location of that network to be at x location, and you search for pizza, they’ll return pizza results near x.
How do they make the “stitching” so “seamless”? It’s clear that at some points one image taken on one day is adjacent to another taken on a different day (you can tell by the different weather). Yet, it seems as though it’s all one picture taken at once. Certainly the car will be in a slightly different position on different days. Does the software match them by coordinates, or simply “intuitive” matching?
buildings have edges and doors and windows, lots of points that can be matched. Trees (in same time period), signs, sidewalks and streets all have points that can be matched.
Note that you can get personal software that does the same sort of photo-stitching (I have one that came with the camera I got a few years ago). It doesn’t work as well as Google’s, but then, Google also has precise control over things like the exact relative positions of all the cameras, as well as the expertise and resources to produce software exactly corresponding to their needs.
This is one of the ways that the iPhone, at least, determines your location - the unique ID of known router locations. In the case of the iPhone, Apple purchased the database from someone else.
The street view of my own house is amazing. I can date and time the photo. Given the fact that there is no snow on the ground, no leaves on the trees, trash cans out and that the sun is shining parallel to the street and that there were no views before last year, I can say that it was about 2:00 PM on Wednesday, April 22, 2009. I can almost read the licence plate number on my car.
