Skip to Main Content

Understanding SmartZone’s Role in Your Forms Processing and Data Capture Workflow

Automated data capture tools are an essential feature of today’s business applications. Without the ability to quickly extract information from incoming forms and documents, organizations will struggle to keep their records, databases, and customer-facing software up-to-date. While software SDKs like Accusoft’s SmartZone can deliver powerful optical character recognition (OCR) and intelligent character recognition (ICR) to help applications accurately capture the information they need, these tools were not designed to operate in isolation. To get the best performance out of them, they need to be incorporated into a comprehensive and well-designed forms processing workflow.

Building an Efficient and Effective Forms Processing Workflow

Although data capture is often the primary objective of forms processing, a number of elements need to be in place for an application to be able to deploy SmartZone’s powerful OCR/ICR functionality. The first step involves the creation of form templates that can be used both for identifying incoming scanned forms and for defining field regions on the page from which data can be extracted. Building this library of templates provides a road map of sorts for the recognition process.

After form images are acquired, either from pre-existing digital documents or newly scanned images, they may need to be enhanced or cleaned up to ensure the best recognition results. Operations such as binarization, despeckling, deskewing, and line removal can all improve the data capture process, especially in the case of scanned documents. Older documents frequently include a great deal of image noise when scanned into digital format, which can make it difficult for an OCR/ICR engine to properly segment and read characters cleanly.

Once a form image has undergone enhancement, it can be matched and aligned with the correct template to ensure that the SmartZone recognition engine will be able to obtain a clean field clip. Scanned images can be overlaid via an alignment algorithm that performs minor adjustments to match it exactly with the correct template. This step is crucial because the data capture process is set up to read the field areas identified by the template rather than recalibrating for each form. If the alignment is off, the engine will not get a clean read of the characters, which could result in inaccurate recognition results.

After the form is identified and aligned, additional enhancement and cleanup operations can be performed on the specific areas of the form that contain information to be extracted. This typically means individual field areas where text or other characters have been entered. The locations to be cleaned up can be designated during the template creation process when data extraction zones are defined. In some instances, a processing workflow may skip the initial full-page enhancement and instead only perform clean-up on areas where data capture will be carried out. This approach is often more efficient from a processing standpoint, especially when targeted, zonal recognition is being applied.

Form image dropout can also be performed at this stage, which involves the removal of image content like signature lines, text field boxes, comb lines, or other extraneous guiding content. Here again, proper form alignment is crucial. If the form is slightly “off” from the template, valuable character content could be removed, making accurate recognition much more difficult. Good form dropout tools should also be able to reconstruct characters that lose pixel data during the dropout process, which is common for characters that have an element that overlaps form lines (such as the lower half of a “j” or a “y,” which might otherwise be read as an “i” or a “v” if not repaired prior to recognition).

SmartZone’s Role in the Recognition Phase of Application Workflows

After a form is acquired, enhanced, identified, and aligned, it can be passed along to the next stage of the workflow for text recognition using SmartZone OCR/ICR. There are a few options that can be selected at this point to help improve recognition accuracy and faster data capture performance.

1. Select Character Sets

SmartZone supports a wide variety of languages and alphanumeric character sets. Realistically, only a few of these sets will need to be used at any one time. Selecting only the sets needed for a particular form will improve recognition accuracy and speed. For instance, there’s no need to have support of Cyrillic languages (like Russian or Greek) enabled if all of the forms being processed are in English.

2. Designate Field Types

SmartZone can designate the expected format of text found in specific fields on a template. Rather than reading each field out of context and extracting the contents without knowing whether or not it’s been filled in correctly, field types can be set to values such as date, email, currency, phone number, or Social Security Number. Regular expressions can also be established for more customizable results. If the character content of the field doesn’t match the designated field type, SmartZone will immediately return an exception and move on rather than trying to recognize and extract the incorrect data. Setting this parameter can greatly improve both accuracy and speed.

3. Set Minimum Character Confidence

Every character SmartZone reads is assigned a confidence value, which reflects the OCR/ICR engine’s assessment of its recognition accuracy. A lower value means that there is a higher likelihood that a character was incorrectly identified. Setting a minimum character confidence value ensures that any character result below that value will be rejected and replaced with a designated rejection character. In practice, this control is used to determine which characters require a manual review following recognition. Setting a high confidence value will ensure higher recognition accuracy, but will likely lead to more exceptions that need to be reviewed by a human.

SmartZone Recognition Results

After character recognition is performed, results can be returned for the character, text line, or text block level. This data can then be passed along to the next stage of a business workflow or used to populate databases connected to the application. Operation instructions, identification, and image areas defined can be transferred to other components for additional forms processing or stored in memory for later access using SmartZone’s Read From Stream or Write From Stream functions.

Getting Started with SmartZone

With support for both OCR and ICR data capture, Accusoft’s SmartZone SDK can serve a vital role in high-performance forms processing applications. The powerful OCR engine can recognize multiple languages, including select Asian, African, and Indian characters. Capable of performing full page or zonal text extraction, SmartZone also includes a variety of customization features that can improve accuracy and recognition speed. Learn more about this versatile SDK’s features and use cases in our product fact sheet.