Understanding OMR, OCR and ICR
In this short tour of the three technologies, OMR, OCR and ICR we take a laook at what the three can do , problems, outlook and in particular what to expect in the gen next OMR.
OMR - optical mark readerThese are extensively used in surveys, election, educational institution.Respondents/examinees answer question by darkening the appropriate circle/oval/box. The difference in light intensity is detected by the scanner and the software evaluates the results and produces reports. Our products use circle. OMR technology is also used in filling of forms. Take a look at a form that reads the name from the darkened circles.
Holy mess isn't it? We hope to make such monstrosities extinct the way dinosaurs are.
OCR - optical character recognitionThis technology is used for digitizing printed documents and books. The document is scanned and each character is recognized and converted into a computer charachter. At the end the printed document is converted into a digital document that can be searched edited and so on. The problems in this technology are due to old documents improperly scanned / photographed documents. These can cause errors in the digitized document. Other than this OCR technology is considered mature and 99.8% accurate.
ICR - Intelligent character recognitionWhile OCR is suitable for recognizing printed characters(in any font) the problem of recognizing hand printed characters is much more difficult. Here is a sample of handwritten numerals (the first 20) from MNIST database Most look reasonable. But not all; for example the second character in the second line.Such characters (outliers) are enough to give the goosebumps to the software.
Obviously the goal of handwriting recognition is to recognize the text i.e. all the words and sentences ICR for words is sometimes referred to as IWR(Intelligent word recognition). Recognition rates on handprinted words can go upto 95% or so. Some have even reported 100% accuracy though on limited data sets. Whatever be the exact status fact remains that in the coming few years we can expect better performance from software in recognizing human handwriting.
Even at present there are systems that tradeoff between high accuracy and manual intervention. If this startegy is adopted it still saves the hassle of manual dataentry in about 90% cases. Cases of manual entry are flagged by software as it recognizes the low accuracy.
What do these developments mean in relation to OMR. If full fledged handwriting with low error rates becomes a reality then OMR would be an antique. However certain things like name and id on our OMR templates can and will use ICR technologies. They would increase the number of questions that can be accomodated and also lead to lesser times for filling up any OMR based forms.