Optical Character Recognition
OCR is used for translating ‘images of text’ into ‘text’. Such text is then understandable by machines, and can be used for further processing. KritiKal has developed a strong in-house OCR engine, which has powered various products and applications like vehicle license plate recognition, container text identification, industrial inspection, document digitization etc. We have experience in designing custom OCR solutions based on client specific requirements; including real time optimized embedded electronic implementations of OCR. The solutions include localization of regions containing text in images, and then using core OCR engine to ‘read’ text out of these sub-images.
APPLICATIONS OF OCR
Traffic & Transportation
License Plate Recognition
Container Number Identification
Vehicle Surface Text Recognition
KritiKal’s Robust OCR Engine helps you increase efficiency across various domains. Here are the features that make it the best in the Industry:
- Robust across different font types, sizes, symbols
- Ability to segment and recognize characters, words, text lines, paragraphs and full pages
- Ability to split and recognize glued characters, as well as to glue and recognize broken characters
- Hi-performance OCR engine, with pre-processing and de-skewing and normalization techniques
Upto 100% accuracy is achievable when the print and image quality is good. With drop in image quality, the accuracy drops gracefully, and can be better than or close to what human beings can read.
200 to 600 Characters Per Second (CPS)is readily achievable depending on the CPU speed.
Color, gray scale or bi-tonal images with a 200 DPI or greater resolution can be used as input. Custom solutions have been built around OCR on videos, integrating frames with varying text resolution (and motion blur).
The output includes ASCII or UNICODE character strings, and confidence values (for decision/data fusion). We have also combined character level output to interpret word or higher level data.
The OCR Engine supports Linux and Windows based platforms. Custom embedded platform implementations have also been made, optimized for memory and run-time.