Log in

No account? Create an account


Millennium Post

« previous entry | next entry »
31st Jan 2006 | 22:01


I just heard a pretty amazing programme about technology and the postal service. The thing that struck me was the way they have mechanized bulk sorting of letters. There are 73 major sorting centres in Britain which between them have 200-300 letter sorting machines, which do the obvious jobs of working out the letter's orientation and photographing it for OCR. (Really fast - 30,000 items per hour per machine.) What surprised me is that the OCR is not done on site, but instead the photos are transmitted over the post office's data network to a single centralized data centre which contains all the clever computers. Of course they aren't so clever that they can deal with all letters, so - second surprise - unrecognized letters are handled by sending the images to offices full of people who type in post codes all day.

No significant sorting intelligence is on the same site as the sorting machines.

Hmm, perhaps this is the manufacturer of the machines: http://www.abprecision.co.uk/businessunits/hsp/postalservices.htm

And perhaps this is a press release about the data centres: http://www.prnewswire.co.uk/cgi/news/release?id=59114

| Leave a comment |

Comments {11}

Tony Finch

from: fanf
date: 31st Jan 2006 23:05 (UTC)

They didn't describe any initial marking stage on the radio, but yes, there is an additional holding loop for letters that can't be OCRed automatically.

Reply | Parent | Thread

from: hsenag
date: 1st Feb 2006 00:15 (UTC)

But without an initial marking stage, how do you re-identify the letter once the postcode is available?

I can see a variety of possible solutions, I'm just interested in what actually happens.

Possible solutions include:
- Rescan letters in the holding loop repeatedly and match up the images
- Treat the holding loop like a TCP window so that you can identify letters in it by position and only need to remove letters from the head

Reply | Parent | Thread