Reading ballot papers using computer vision
Nerrida Dempster · The University of Queensland · 1999
Currently in Australia information from electoral ballot papers is obtained by individuals who sort and count each paper by hand, and then type this handwritten information into a computer system. The process is laboriously slow, expensive, and potentially erroneous. A system that optically scans and interprets handwritten data on ballot papers, and inputs the information directly into the existing computer system would save time and money, and increase its accuracy. Software for a scanning device that fulfills this aim has been developed. Full integration with the Australian Electoral Commission’s computer system is required, along with further developments in scanning hardware before this system can become operational. At that time, a fully automated ballot-reading system that is accurate, reliable and has a fast processing time will be possible.