An optical scanning device combines specialized computer hardware and software. The hardware devices capture an image and software converts the image to computer-readable data.
Voters using machine-readable ballots are given a ballot card with the names of candidates printed on it. Next to each candidate a symbol is printed, such as a rectangle, circle or incomplete arrow. The voter indicates a choice for a candidate by filling in the appropriate rectangle or circle or by completing the arrow.
After voting, the voter may feed the card directly into a computer vote tabulating device at the polling place, or the voter may place the card in a ballot box, which is later transported to a central location for tabulation.
The computer tabulating device identifies the marks made by voters on the cards and records votes accordingly. The individual votes are recorded in a database and aggregated to give total results.
Marksense systems were used by 24.6% of registered voters in the U.S. 1996 Presidential election. Use of these systems in the U.S. is increasing as older lever and punchcard systems are replaced.
There are four main types of optical scanning technologies:
- Optical Mark Reading (OMR)
- Optical Character Recognition (OCR)
- Intelligent Character Recognition (ICR)
- Imaging technology Optical Mark Reading (OMR) scanning systems
Most machine readable/optical scanning voting systems use OMR technology. OMR technology has been widely used since the 1970s for a variety of purposes, including school and university tests, censuses, surveys and lotteries, as well as for voting. It is also used in barcode readers, which are in widespread use in retailing, stock taking, libraries and schools.
OMR typically involves a scanner reading particular kinds of marks in a defined set of locations on a page. The computer software used by the OMR scanner is programmed to recognise the meaning of the various marks and to convert scanned images into computer-readable data using the location of those marks.
OMR systems are well suited to first-past-the-post and list electoral systems, where voters are asked to make simple choices when voting, easily represented by a simple mark. In more complex electoral systems, such as alternative voting systems and single transferable vote systems, where voters are asked to choose candidates by showing sequential preferences, it is more difficult to apply OMR technology. As a result, scanning technology has not been used widely for counting these kinds of ballots. However, the increasing accuracy of ICR may make it a viable technology for these kinds of ballots in the 2000s.
Apart from voting systems, there are other potential applications for OMR technology. In Australia, for example, OMR systems are used to scan electoral rolls marked in polling places to indicate the names of electors who have voted. This permits Australian electoral authorities to automate the enforcement of Australia's compulsory voting system, as well as identify any instances of multiple voting.
OMR barcode systems are also used extensively for mailing applications. Many postal authorities place barcodes on mail to automate the delivery process. Some countries have mail systems that allow users to print address barcodes on mail so that postal authorities can process the mail without having to print barcodes in their mail-rooms, thereby creating a discount for the user.
Electoral authorities are also making use of barcodes on mail. Since barcodes can identify both the name and address of the voter, they can be used by electoral authorities to process the mail when it is returned. This is particularly useful for postal ballots to automate the recording of voter names. These barcodes can also be used where addressed mail is returned “not known at this address” in order to capture those details for electoral roll update purposes.
OMR technology is very useful for and efficient at gathering relatively simple, pre-determined data. However, it is not very good at gathering complex, variable data, such as large amounts of text. OCR and ICR systems are more suited to this purpose.
Optical Character Recognition (OCR) scanning systems
OCR scanning systems take scanned images and use computer software to recognise the shapes of printed or handwritten characters such as numbers and letters and store them as computer-readable data. OCR is typically used to convert printed text into computer-readable text.
This capability has many potential applications in the electoral field. For example, in the early 1980s, the Australian Electoral Commission produced an extensive set of procedures manuals. Some years later, when the manuals were due to be revised, the original computer files containing the manuals were not able to be used by the Commission's upgraded computer software. Rather than retype the original manuals, OCR software was used to convert the printed manuals into computer files suitable for editing and revision.
Another important use of OCR is for data capture of information printed on forms. Rather than manually typing information contained on forms, OCR can be used to automatically convert information from forms into computer-readable data.
OCR works by “training” the scanning software to recognise particular shapes as letters and numbers. Since different print fonts are different shapes, OCR systems have to be trained to recognise that a particular letter or number can take several different forms. Given the regularity of printed fonts, this is a relatively straight forward process. OCR systems can also be trained to recognise hand writing. However, given the infinite variety in hand writing styles, this is a much more difficult task.
Early OCR systems had a relatively high error rate when converting printed text to computer-readable data, particularly hand written text. This required a high level of human intervention to proof-read and correct the converted data. As optical scanning hardware and software improved towards the end of the 1990s, the error rates dropped. However, the next generation of scanning systems, ICR systems, went even further in increasing scanning accuracy rates.
Intelligent Character Recognition (ICR) scanning systems
ICR takes OCR systems one step further by using computer software to apply intelligent logic tests to scanned characters so as to more reliably convert them into computer-readable data.
ICR systems apply rules of spelling, grammar and context to scanned text in order to make “intelligent” assessments as to the correct interpretation of the data. This enables much more accurate conversion of scanned text than does the more simple OCR system, particularly with handwriting.
ICR software requires fast, powerful computers to perform efficiently. Reliable ICR systems only became available in the mid to late 1990s with the development of cheap, powerful computer products.
As ICR systems become more reliable, their use for electoral applications will increase. They are particularly suitable for capturing data from forms. ICR systems are also being examined for their suitability to capture hand written numbers from ballots used for more complex electoral systems, such as alternative vote and single transferable vote systems. To date, automatic data capture systems have not been used for these electoral systems owing to the complexity of the task.
Imaging technology
In addition to capturing images for conversion into data, scanners can also capture images to be stored as computer-readable images. Photographs, drawings and images of text can be stored and reused in computer-readable form.
Computerised images have many electoral applications. Images can be included on websites and printed in publications. Staff photographs can be placed on an electoral authority's “Who we are” Internet page and in its Annual Report. Photographs of polling stations can include on websites and instruction manuals. Examples of completed forms can be scanned as images and printed in training manuals.
Paper-based forms can be imaged and stored in electronic form. Copies of the images can then be downloaded over a computer network without the need to access the original paper copies. The Australian Electoral Commission is currently engaged in imaging all of its millions of voter registration forms and placing them on a computer network accessible from any of its offices nationwide. This system will be used to check signatures or any other details included on the forms by accessing the imaged forms on demand.
Corporate logos can be stored electronically as images and printed on a range of corporate publications. Where an organization may once have used expensive pre-printed stationery containing the corporate logo, stationery can now be printed from the desktop with professional letterhead using blank paper, a colour printer and a digitized image of the logo.
Imaging technology can also be used for identity verification purposes. Photographs can be digitized and placed on identity cards. Images of finger prints or facial features can be digitized and stored on smart cards. Software identity systems can be used to compare the image of the person presenting a smart card with the image of the person encoded on the card to determine whether it is the same person.
