Image processing and quality control
Image quality is a critical factor in determining how well characters and documents are identified and processed.
Automated data capture software includes advanced image enhancement and correction capabilities, including de-skewing, auto-rotating and character enhancement.
Built-in quality control features identify and correct low-quality images early in the process to reduce human intervention.
Automated document recognition and classification
Each image received will be compared to one or more pre-defined templates to identify the form type. The software can recognise and distinguish between many different form types and process them accordingly. Unrecognised pages (such as cover pages or damaged pages) will be routed to an unclassified queue for manual handling.
Extract data from scanned documents using OCR, ICR, OMR and barcode recognition
After document processing and classification, data capture software uses handprint (ICR), machine print (OCR) and checkbox (OMR) document recognition technology to automatically extract data from scanned documents. This will include constrained print fields, tick boxes, variable scales, comments, tick boxes and signatures.
Intelligent character recognition (ICR)
To read short hand-print written responses from constrained print fields (one character per box) such as names, dates and numbers.
Optical mark recognition (OMR)
OMR technology identifies if a checkbox has been filled, with automatically handling of crossed out and amended responses.
Optical character recognition (OCR)
To capture machine printed text.
Image zone capture
Predictive key from image, coding and image snippet capture can be used to capture drawings or cursive handwriting.
To recognise all standard barcode types, including 2D matrix barcodes.
To confirm if a paper form has been authorised with a signature by calculating the fill percentage of the field.
Simple rules such as alpha, numeric, dictionaries, date ranges, look-ups and mandatory fields will be checked at this stage with any unrecognised fields/characters queued for human review.
These common-sense logic rules are applied to the extracted data to ensure that invalid responses are not exported (e.g. impossible to tick more than one response for a single-choice question). In such situations, the exception will be intelligently routed to the right human operators to review and correct. The entire process takes seconds meaning thousands of forms can be processed each day.