White Paper

US Census Demands Reliable Imaging Solution

The Problem
In 2000, the U.S. Bureau of the Census (BOC) processed more than 120 million household surveys. Considering that forms are filled out by hand, the BOC required a document imaging system that could recognize inconsistent markings, accommodate illegible entries, convert the forms into electronic data files, and do it in half the time it took the prior census in 1990. The project's prime contractor, Lockheed Martin, didn't have to go far in its search for a solution. What's more, it didn't have to invent new technology to perform at high levels of accuracy and speed. Lockheed Martin turned to Kofax Adrenaline scanner accelerator boards and ImageControls software, a cost-effective product combination that delivered a successful outcome.

The Solution
By implementing a data capture and management solution featuring Kofax products, the census was able to process more than 154 million documents in less than 170 days. Never before had the bureau converted this much data in such a short period of time.

The BOC's key objective in 2000 was to employ technology to quickly and accurately count and evaluate completed census survey forms. Lockheed Martin Corp. was contracted by the BOC to implement an imaging and recognition solution to complete the processing of census surveys within 171 days. The company assembled a team of leading imaging and recognition, as well as document management technology vendors which included scanners from Kodak, Adrenaline scanner accelerator boards and ImageControls toolkit from Kofax, workflow software from Staffware Corp., image processing software from TMSSequoia Inc., optical character recognition software from CGK, data correction software from Captiva Software Corp. and optical mark recognition software from Optimum Solutions Inc.

Taking the U.S. Census has never been easy. Back in 1790, the first Constitutionally mandated census was ordered to count the 3.9 million people free of English rule and taxes scattered throughout the original Eastern states and territories. The "technology" involved door-to-door canvassing by U.S. Marshals, a tedious and time-consuming process challenged by an ever-expanding frontier. In just 210 years, the country's population has increased by more than 70-fold. The 2000 U.S. Census counted 284.6 million residents inhabiting the United States and its territories. It also was the first U.S. Census to use digital imaging and automated recognition technologies with the assistance of document capture hardware and software products from Kofax.

Interestingly, the methodology for taking a national census every 10 years is as old as the U.S. Constitution itself. Then and now, survey forms are distributed to households and census workers visit homes to interview residents. Originally, tabulations were done by hand. The first rudimentary punch-card tabulating machines from IBM, NCR and other companies were developed in the 1880s. With the new millennia, however, the BOC opted to introduce technology to a process that emerged from an era of patriots, plantations and frontier expansion.

Kofax 1700 Adrenaline boards were paired with Kodak 9520 high-speed scanners to convert the paper-based census surveys into electronic data. The Kofax ImageControls toolkit was used to provide both a scanner interface and advanced barcode recognition algorithms, allowing the system to quickly process the census forms. The Kofax-enabled forms-processing technology played an integral part in the BOC's Data Capture Center (DCC) operation, which operated in four regional offices across the nation. The DCC operations were managed by TRW. Each DCC office operated a cluster of integrated technologies which included scanners, workstations and servers to gather and convert the region's census information into electronic data. The offices transferred the data on a daily basis via high-speed T1 private network telephone lines to a central database at the Census Bureau's headquarters in Washington D.C. for review and evaluation. Each DCC processed at least 35 million forms, with some DCC's processing over 40 million forms.

"The use of imaging and recognition technology enabled the BOC to reliably extract the hand-printed information from the census forms and record it as ASCII data," said Bill MacDonald, Lockheed Martin deputy program manager for DCS 2000, U.S. Census Program. "The system also enabled census workers to manually review and correct data that could not be accurately read by the document capture system to exceed the bureau's 98 percent accuracy objective."

The BOC's DCC locations in Baltimore, Phoenix, Pomona, Calif. and Jeffersonville, Ind. were required to process all forms within two days of receipt. During the peak mail-back periods of mid-March and mid-April, more than 6 million forms were received each day. This was equivalent to 64 tractor-trailers of documents. The human component was immense: Each DCC office was staffed by 2,000 census, TRW, and Lockheed Martin employees working two shifts 24/7.

Solution Benefits
"The BOC recognized that document imaging technology could complete the job of processing census forms more effectively than manual hand counts," said Doug Rudolph, Kofax, vice president of sales for the Image Processing Business Unit. "With the unique features and powerful barcode recognition capabilities of Adrenaline and ImageControls, the Census Bureau was able to capture data faster and more accurately than any previous census in history. This enabled the BOC to complete its task of processing returned forms two weeks ahead of schedule, allowing the census to get an early start on sending workers out to interview residences that did not return completed surveys."

The methodology and technology for taking the census have traveled a long road since the Revolutionary era and the dawn of digital computing. Hand-entry ledgers and bulky, noisy punch-card tabulating machines have given way to high-speed image capture systems and rapid forms recognition technology. The result: Census forms are now scanned and processed with greater than 98 percent accuracy. Ultimately, the people greatly benefit through a fair apportionment of seats in the House of Representatives and the equitable allocation of billions of dollars in tax money throughout the nation.

Kofax