In the first Ten Minute Tip in this series we saw how to use EXIF data to geolocate an image. Unfortunately most images found on the internet have their EXIF data removed, so this approach is not always possible.
However we also began to use a three-step methodology to geolocate images, and we can apply this even when there is no metadata to help us out. By extracting, researching, and then verifying the data in an image it’s still possible to find out exactly where an image was taken.
Extracting The Data
The image in this geolocation task contains a lot of data that will make it possible to find exactly where this image was taken. Before that happens, we need to extract all the information from it that we can. This time instead of using software to extract metadata, the information we need will be extracted with nothing more than our own eyes.
So what can you see in the photo? It’s important not to try and assess the image from an aesthetic viewpoint. In geolocation each image has to be treated as a repository of raw data that needs to be picked out and sorted. This means that we need to look not only at the central parts of the image, but all the peripheral details too – what’s in the background? What does the blurry sign say that’s almost out of shot? Which details matter to us even if the original photographer didn’t focus on them?
A useful method to make this easier is to write down a list of twenty pieces of information you can see in the image. Sometimes extracting this many data points can be hard, but it forces you to be thorough. The more time and effort spent extracting information from the image will make the research and verification stages much easier.
Is it possible to extract twenty pieces of information from this image? Let’s try.
- Sign: “Halls Ish & Chips” (probably Fish and Chips)
- Cyclist picture on front of house
- Flag in the window – white flower emblem
- Road sign “Masham B6268”
- Green truck and yellow digger in the background.
- Red and white signs – one possibly says “SOLD”
- White digger with “Cemex” logo
- Email address: “estimating@cemex[.]com”
- Multiple roadworks signs
- Pennants in the window – same as the flag in upper window
- Roadworks permit number
- Street name: “Market Place”
- Sign: “Boutique Cafe”
- Number: 42A
- Sign: “Bedale Leisure Centre”
- Sign: “Thorp Perrow Arboretum”
- Sign: “Bedale Hall”
- Sign: “Wensleydale”
- Sign: “Camp Hill Adventure Park”
20. Sign: “Bedale Camping & Caravanning Park”
That’s probably going to be enough information to start with, so that’s it for the extraction phase. Next we can start to research some of the extracted information to try and find out exactly where the photo was taken.
We could start our research with any of the extracted information, but it’s already clear that some points might be more useful than others. As a general rule, information about permanent features like street names, addresses, and architectural or geographic features are a better place to begin than temporary features like roadworks, passing traffic, or other public events. Temporary events like these can be useful for identifying when something happened, but we need to know where first.
For this image there are a lot of indicators that include the town name “Bedale”, so it’s a good place to start researching. Here’s the result of a simple Google Maps query:
There’s a little town in England called Bedale, but before jumping to conclusions we can add in some of the other details we found. We can check if Thorp Perrow Arboretum is nearby (point 16).
It looks like it is. So this means we’ve probably got the right town – but we want to find the exact spot where the photo was taken. To take things further we can go back through any of the data points we initially extracted and add them into the research.
We alreadyhave a street name of “Market Place” (12) and a house number “42A” (14). Let’s see if they can help us refine the location searching:
That looks pretty good! Google tells us that there’s an “Institution Boutique Cafe” there too, which matches point 13. Before moving to verify the exact photo location, it’s important to underline a point about the methodology.
The presence of a road sign makes this particular image easier to geolocate. Why? Because it offers a lot more information to extract and work with.
But what if the road sign wasn’t in the picture? It would look like this:
We could still use the same method to come to the same conclusion about the probable location. If we don’t have road sign data to work with, we could do a Boolean search for “Halls *ish & chips” like this:
There are only a handful of possible matches – and the first one leads to the exact same location in Bedale. Road signs really help, but they are not essential to find out where this was taken.
So far all the research points towards the location being in the middle of the town of Bedale, but we need to verify this is the case and never simply assume that our research doesn’t need to be checked. This location is covered by Google Street View. Street View is enabled by clicking on Peg Man:
Then click on a blue line to see street level imagery in the location of interest. Here’s the view from outside Halls Fish & Chips:
There are no roadworks of course, because they are a temporary feature that was not present when the Google Image was taken. There are a few other minor differences because this image was taken at a different time to the original. Nevertheless the street names, building numbers and business names are the same. The layout of the streets and buildings is also exactly the same, so we can be confident that our research has led us to the right place.
You can see that it isn’t necessary to have EXIF data to find out where an image was taken. By spending time carefully extracting, researching, and verifying the information in a photograph it’s still possible to geolocate it with a high degree of precision.
In the next tutorial we’ll step up the difficulty again and find out how to apply the same methodology to successfully geolocate images that have much less detail in them.