Bar codes were invented in 1962 to enable automatic tracking of railroad
cars. Early in 1970, they were first used on grocery cartons, and a
few months later, for automobile components tracking. Widespread adoption
of modern bar coding occurred because it was simpler and easier to read
than optical character recognition (OCR) technology,
and it guaranteed accuracy - the same features now driving the use of
bar codes in document imaging.
Using bar codes in high-volume document imaging systems to automate
indexes and document control can improve user productivity dramatically,
while providing major benefits over other recognition technologies.
Following a few simple guidelines can ensure the successful addition
of bar codes to your customers' document imaging applications.
A number of creative bar-code applications are emerging in conjunction
with document imaging.
In the transportation industry, bar coded tracking numbers are affixed
to invoices or delivery documents, marked with notes on shortages or
damage. It's critical for a distributor to receive this data promptly
so a customer service department can react.
In distribution or parts warehouses, bar codes can be affixed to a
page as parts are used. The paper can be faxed and the image of the
bars converted and input into an inventory control system.
Pharmaceutical companies bar code medicines and prescriptions. These
can be faxed from a pharmacy to a doctor's office each time a patient
fills a prescription, allowing physicians to automatically update patient
But beyond these implementations, the two main uses for bar codes in
document imaging are batch scanning, such as on batch control sheets,
and automatic indexing.
Batch scanning is being used more and more in high-volume applications,
defined as 5,000 or more sheets of paper to scan per day. The simplest
and most common method for scanning paper is to place each single or
multi-page document into a scanner's hopper, scan and view each page,
and enter index fields from the keyboard. However, working this way
cuts productivity, because each scan takes about 20 seconds, by the
time indexes are located and keyed, compared to the two or so seconds
typically needed to scan the sheet itself. Batch scanning lets users
make up the lost time.
There are two methods of batch scanning. One is used when similar documents
are being imaged repeatedly, such as with Federal Express forms processing.
The second is used when a group of pages represents an indexing entity,
such as a medical claims form followed by supporting documents.
Batch scanning requires the creation of batch control sheets as the
front of each batch. These may be used for manual control, or may be
encoded so the system can recognize and control the batch automatically.
A second, more complex type of batch control sheet has a meaningful
bar code with a batch number encoded i\onto it, or in the case of litigation
support systems, a case number. The software determines the start of
a new batch by locating the bar and reading the batch number from the
Alternatively, a bar code can be placed on the outside of a folder.
In this case the operator passes a wand over the bar before stating
to scan the batch, assigning a batch or case number to attach to each
image in the batch.
Barcodes have been developed in a number of differing formats. The
oldest now in use is Codabar, developed in
1972 by Pitney Bowes and stills used for Federal Express airbills, as
well as in blood banks and libraries. Most document imaging systems
seem to have started with Codabar. It's easy
to read and relatively large; also, a clear 14-point interpretation
printed above it can be verified using OCR.
Interleaved 2 of 5 , the second-oldest bar
code, is used on corrugated cartons and in the pharmaceutical industry.
If possible, avoid using these codes in new applications, because their
character sets are limited and can give false reads.
Code 3 of 9 , was adopted by the Department
of Defense in 1981 and has become the prevalent code among industrial
applications and in the transportation industry. It is reliable and
flexible due to a checksum and support of both alphabetic and numeric
characters. Its main drawback: it uses a lot of label space, and therefore
canít be used on small documents. For this reason, it's also unsuitable
for printing many codes on a single sheet.
Code 128 , announced in 1981, solved many
problems associated with the previous codes. It supports all ASCII
characters , uses minimal integrity due to several separate message
check routines, Its only drawback for document imaging is that it uses
proportionally sized bars that can prove difficult to interpret, particularly
UPC, the retail trade bar code, is rarely used in association
with document imaging, nor are Postal bar codes.
Automatic indexing using bar codes is becoming popular because OCR,
the most obvious technology to use, frequently fails to deliver accurate
conversions in typical forms-based document imaging applications. The
forms get in the way, numbers and characters often are skewed, and documents
are faint and sometimes dirty. Backgrounds and color interfere with
the image, and it's difficult to locate the key fields automatically.
As a result, OCR usually achieves successful recognition less than 90
percent of the time, leading to expensive index repair keying.
Bar codes solve all these issues. Software to locate the distinctive
pattern of black and white lines is available. Provided that transitions
and thicknesses can be identified, bar codes can be correctly converted;
check digits will guarantee the conversion. With a document imaging
system that can deskew, bars can appear at any angle.
A 200- dpi scan equals .005 inches per pixel. Since the thinnest bars
in any bar code scheme, known as the X dimension, are usually printed
from 10 mils to 0.05 mils (i.e., 0.01 of an inch to 0.05 inches), the
thinnest lines normally will contain 2 pixels across. Federal Express'
Codabar scanned at 300 dpi, for example, consists of wide bars and spaces
of about 14 to 16 pixels wide, and thin ones of about 5 to 7 pixels
wide. It's not sensible to reduce thin line width below 2 pixels because
it may cause conversion errors.
However, a faxed bar code transmitted at 200-dpi horizontal resolution
can be accurately converted. Paper or bar code skew is less of a factor
than with OCR. If the program can effectively draw a straight line through
the bar and see the differentiating black and white pixels, the decode
will be accurate. Be aware of whether the bar code is vertical and the
fax is in normal mode, because the scan rate is 200 dpi horizontally,
but 100 dpi vertically. Due to the success in interpreting faxed bar
codes, a number of applications have been developed in which bar codes
are sent over the fax.
Achieving maximum efficiency
Although bar codes are easy to read and implement, there are guidelines
for achieving peak efficiency. As we've discussed, not all bar codes
are created equal. All bar codes doe have a fixed number of bard per
character, so if a bar is missed, the code becomes unreadable. If adjacent
bars are touching, it' possible the bar code's size will be misread;
but because the table of bars is smaller than the number of possible
combinations, an illegal character will be generated and then ejected.
As a result, bar codes are significantly more secure than OCR. Substitution
is virtually nonexistent and checksums will eliminate errors. Barcodes
can be read in either direction and contain a different start and stop
code. This can be used to recognize when a bar code is upside down,
allowing a page with a preprinted bar code on it to be automatically
Alternatively, if the barcode was stuck on in a warehouse, it might
have been accidentally placed upside down. If software can recognize
these situations, it can deal with them.
The higher the scan resolution the higher the bar code density can
be, although the user must ensure that the bars are not close together
that they touch; again, a 2-pixel width is the lowest you should go
to ensure accurate reads. Poor print quality from a dot matrix printer
or highly skewed images will increase the probability of lack of scan
separation between the bars. Ribbon quality must be closely monitored
when using a dot matrix printer to create bar codes.
Most imaging systems scan at 200 or 300 dpi, so the size of the bar
code and the density is restricted. This affects the amount of information
held on the bar code. Locating the bar code within an overall image
can be difficult. Many products have hard-coded the search location
of the bar code. This technique is acceptable for predefined batch header
sheets and preprinted bar codes, but lacks flexibility. Even if the
bar code is always in the same place, a page may be upside down with
Using a laser printer is an easy way to generate bar codes. Many software
packages allow bar code printing from PCs. For higher-volume applications,
specialized bar code printers that can churn out multiple stick-on labels
are available. Beware of dot matrix printers for bar code labels, unless
you have space for a very large bar doe such as those found on batch
Picking the right bar code for document imaging
If a standard bar code format is not in use and you need to select
a bar code for document imaging, research the following before looking
for a solution:
- Is the required date numeric only or alphanumeric, with or without
special characters, and will it expand in the future?
- How much space is available for the bar code, and how many characters
- Whatís the possible position on the document? Some bar code technologies
are more tolerant in areas such as reading labels placed at the edges
- What resolutions will you scan at, or what is the delivery mechanism?
- How many misreads can you accept?
- What's the maximum number of bar codes that you are going to have
on the page?
- What methods will be used to print bar codes, and in what form will
labels be issued (i.e., preprinted, stuck on, etc.). In general, select
the simplest and largest bar code an application will allow. A quiet
zone or blank area of 1/2 inch or more before and after the bars is
helpful to improve accuracy. Print the translation above or below
the bar code in 10+ point font to ensure readability for verification.