The USPS uses machines that recognize numbers and then letters.
Neither snow nor rain nor scribbled zip code stays these couriers and their computer counterparts from the swift completion of their appointed rounds.
The United States Postal Service (USPS) began researching remote computer readers (RCRs) for handwritten addresses in 1983. At the time, the technology required to scan and understand a human scrawl simply did not exist.
Not until Christmas of 1997 did the USPS and the University of Buffalo's Center for Excellence in Document Analysis and Recognition (CEDAR) deploy its first handwritten address-reading prototype, which rejected 85 percent of envelopes and correctly identified the address in only 10 percent of those it read with a 2 percent error rate.
Humans read and comprehend with an ease that belies the immense difficulty of computer pattern recognition (including patterns such as numbers and letters). It is one of the central problems in artificial intelligence. So even the prototype's 2 percent was viewed as a great success.
"That Christmas alone we saved several thousand dollars for the post office," Sargur Srihari told LiveScience. Srihari founded CEDAR and led the early research on large-scale RCRs.
Today, the large majority of letters sent through the post office are read and sorted entirely by computer. According to Srihari, current reading success rates are above 90 percent.
The process begins when a single envelope is scanned into a digital image. Initial algorithms locate the destination address on the envelope (as opposed to the return address), then parse this address block into discrete lines, and those lines into words. Finally, the software is left with the street number, street name, city, state and zip code.
Actual recognition begins with the zip code and street number. "Numbers are easiest to read," Srihari said, "because they are usually well-formed compared to letters, and there's only 10 numerals."
The program takes these number-values and searches through a database of about 120 million U.S. addresses, emerging with a short list of possible candidates. Then, the computer compares the individual letter features of the handwritten street name to the candidates. "It's a tricky business," Srihari said, "but you can usually recognize a few key features [of the address words]."
"Each word gets a confidence rating. We go through all possible street names, and find the highest confidence value."
The letter is then shipped, first to the correct city, then to the nearest post office. There, the letter is scanned by computer and filed into the appropriate position in a carrier's stack, so that the first human eyes to examine the envelope are those of the postal carrier approaching your mailbox.