Optical Character Recognition has traditionally been performed using a 2 step process in HALCON: First the characters in the image are segmented from the image background. Next the characters are read using one of HALCON’s classifiers. Of these two steps, the most challenging was usually segmentation. In HALCON 10, MVTec introduced an operator that could automatically segment characters, but to do so accurately required determining correct parameter values such as character height and width, mean stroke width, text structure, and polarity (light on dark background or vice versa). For certain applications this was not always easy, and in some cases, the solution was not generic enough to apply successfully from application to application.
In HALCON 12, MVTec introduces the brand new “automatic text reader” that can handle variations in character features, is easy to use, and can both detect and recognise characters in a single operation (create_text_model_reader). The new automatic text reader provides a generic solution for a wide range of applications and conditions. It is able to detect text with different font sizes, polarity, varying backgrounds, and with slight orientation variations; all without the need to set a single parameter. Adding to the improved performance, MVTec introduces an enhanced MLP classifier which offers more robust recognition of segmented characters.
The new automatic text reader has proven to be very successful across a wide range of images and situations, however there are a couple limitations: It cannot currently read dot-print font or imprinted characters with reliable success. For these styles it is recommended that the manual OCR method be used. Additionally, the automatic text reader can’t read specialised fonts like hand writing or artistic fonts, and it can’t find single characters.
The automatic text reader in HALCON 12 provides developers with an easy to use OCR tool with excellent performance, and it significantly reduces development time. To view example programs of the automatic text reader in action, download HALCON 12 or contact sales@multipix for a free evaluation licence.