By Thomas J. Hartley of Palo Alto, CA
I joined Philco’s Corporate Research Lab in July 1959. In January 1960, our lab received funding from Philco’s Computer Division to begin development of an optical character reader (OCR). The Computer Division wanted a machine that could read data from the computer’s high-speed printer.
The experimental OCR we implemented could be programmed to recognize any one alphabet character. We used a flying spot scanner imaged onto a flat surface, where a document could be placed, and an array of photo sensors to detect the light reflected from the document. The flying spot rapidly scanned from the top to the bottom of a line of print, then slowly scanned across it.
The video was quantized, and the resulting binary stream was fed into a shift register. In effect, the image captured in the shift register slowly scanned across the line of print. An electrical output from each element of the shift register was wired to a rectangular patch panel laid out to present the data as it would appear in the document. We had a large number of resistors molded into plastic and fastened to patch panel plugs that we could insert into the panel to create a weighted template for recognizing a character. The currents flowing through the resistors were summed and compared to a threshold value. If the total current exceeded the threshold, then the character was recognized. This was a simple implementation of template matching.
My responsibility was to supervise the technicians who assembled the OCR and designed recognition logic for specific characters. My first design was for the letter “P” as in Philco. The OCR immediately became known as the P-Picker.
Not long after I completed a successful “P” design, representatives from the Post Office Department were brought in to demonstrate. After we showed them how to pick out the “P” in a line containing the complete alphabet, I was asked if I could have it recognize the letter “R.” I asked for some time to put the logic together, and they went off to meetings. When they returned, I successfully demonstrated the ability to detect the “R” without falsely detecting the “P” in the same line of printed characters. Not long after, Philco was awarded a sole-source Post Office Department contract to develop an OCR to read addresses on envelopes.
As a direct consequence of the P-Picker demonstration, Philco developed and delivered 30 postal OCRs, which were installed and put into operation in major Post Offices around the country. These machines were capable of reading the city and state addresses on machine-imprinted envelopes and sending the envelopes, along with the destination information, to a large mail-sorting machine. One year, they were credited with sorting over 4 billion pieces of mail with greater accuracy than human operators who entered envelope destination information via keyboard.
Editor’s Note: Thomas J. Hartley was a Life Senior Member of the IEEE. He has a BSEE and MSEE from University of Pennsylvania. He joined Philco Corporate Research Labs in 1959. he made significant contributions to the first practical implementation of Optical Character Recognition and image pattern recognition. He died in 2016.
An OCR II system developed by Philco-Ford was later installed in Boston’s South Postal Annex. This machine was notable for reading the entire address, including the house number or box number and the patron’s name.
