By 
Google introduced street views to selected areas of their maps recently.
Not discussed in Google’s Street View help sections is how those views might help Google improve the accuracy of locations for both the maps, and for Google’s business locations databases.
A problem with search engine mapping databases is that the information collected hasn’t always been very accurate, based upon the way that the this kind of information has been collected.
Often, GPS location information for some “anchor” street addresses are known for these systems, and the locations for businesses and buildings between the anchors has been interpolated.
According to the authors of a new patent application from Google, significant discrepancies are sometimes observed between actual GPS locations and interpolated locations, with actual addresses being off by 100 yards or more.
So, how might the Street Views help?
The patent application is:
Database assisted OCR for street scenes and other images
Invented by Bret Taylor and Luc Vincent
US Patent Application 20070140595
Published June 21, 2007
Filed: December 16, 2005
Abstract
Optical character recognition (OCR) for images such as a street scene image is generally a difficult problem because of the variety of fonts, styles, colors, sizes, orientations, occlusions and partial occlusions that can be observed in the textual content of such scenes.
However, a database query can provide useful information that can assist the OCR process. For instance, a query to a digital mapping database can provide information such as one or more businesses in a vicinity, the street name, and a range of possible addresses.
In accordance with an embodiment of the present invention, this mapping information is used as prior information or constraints for an OCR engine that is interpreting the corresponding street scene image, resulting in much greater accuracy of the digital map data provided to the user.
The method used involves a mix of using a business location database, optical character recognition, and panoramic street view images to better map out the locations of businesses shown in those streetview images.
A process for using OCR with street scenes may involve:
- Getting a target GPS location for an image in a street scene using known GPS data associated with the image,
- Estimating a street address of that GPS location,
- Determining a range of addresses shown in the image, based upon that street address,
- Running a query on the mapping system database to find a business name having a street address in the target address range,
- Using OCR on the street scene image to see if key words associated with that business name are present and;
- If one is, determining an actual GPS location for the street address of that business name, based on the known GPS data.
A visual analysis of the street scene image could also take place to see if expected features associated with the business are present.
The OCR method described could identify both textual and non-textual features, to see if non-textual expected features are present. Examples might be things such as golden arches for a McDonalds or a red roof for a Pizza Hut.
One way Street Views could prove useful is simply in being able to see what a street really looks like when you’re using the map for directions. I think most people are fine looking at a map and understanding which way to turn to get where they want to go, but there are some who prefer the visual. Think of the people who give you driving directions and how some will tell you to turn on a given street where others will tell you to turn at a specific landmark.
The idea of being able to search for specific businesses along the street based on an image of the street is certainly a useful feature. The technology might not be there yet, but in time it will.
There is a lot of value to being able to identify a location that you are looking for by recognizing it by the street view image.
I like the approach that Google described in a patent application on Human centric driving directions, which can benefit a lot from the use of street views.
[...] As if Google Street View wasn’t cool enough Bill Slawski has found a patent application indicating that Google might be able to use Optical Character recognition or OCR to pinpoint the locations of thousands of businesses across the world. [...]
[...] A potential future source of address information might come directly from signs on buildings in the street view images of the buildings, which I wrote about in Google on Reading Text in Images from Street Views, Store Shelves, and Museum Interiors. Another post on that topic is Better Business Location Search using OCR with Street Views. [...]
[...] they have an advanced geo-OCR application that they are using to analyze & interpret their StreetView image data and convert the road signs etc.(they spoke about this briefly in their [...]
[...] can read. That really shouldn’t have been a surprise back then. I wrote a post in 2007 titled Better Business Location Search using OCR with Street Views which described how Google might use OCR to gather information from signs it takes video of for [...]