Europe, UK, Asia & RoW: [email protected] +44 (0)1622 859444 | USA & Canada: [email protected] +1 401-583-7846

Decoding the Future: The Role of Optical Components in OCR

12th Aug 2024

Optical Character Recognition (OCR) is set to witness huge growth throughout the next seven years [1], with its market value estimated to reach $39,785m by 2031 [2]. With several OCR variations on the market designed to improve and automate an array of functions, these setups are now integral for an assortment of fields – whether it’s sorting letters and parcels for postal services – such as Royal Mail and TNT Post – and stock taking in retail contexts to inspecting best-before dates, serial codes, and vehicle identification numbers (VINs) in manufacturing. In this blog, we explore the latest innovations in OCR, its utilisation, and the essential optical components powering them.

A picture of a camera lens

Boosting Business Efficiencies with Commercial OCR 

As more corners of the world adopt digital transformation and move practices online, the requirement for OCR platforms and their affiliated machinery has grown considerably. Hardware typically comprises optics-incorporated tools like cameras, scanners, imaging devices, and smartphones. With many businesses transitioning online, there is a notable increase in offices using digital scanners to digitise large quantities of paperwork promptly. As well as digital scanners, you’ll also commonly find flatbed versions and smartphones being used to scan smaller volumes of documents and individual sheets of books and files. 

A man scanning a sheet of paper with a smart phone

Companies use this strategy to quickly obtain details – containing both printed and handwritten text – from items like posters, drawings, product labels, articles, reports, forms, and invoices.

In conventional scanning like this, a range of optics are used, such as:

OCR Solutions for Reading and Preserving Historical Documents

Alongside business documents like invoices, reports, and forms, the optical method also plays a crucial part in reviewing and conserving historical archives all around the globe. OCR is continually improving, helping us gain a deeper understanding of the bygone eras. For instance, in 2018, The National Archives highlighted how OCR – which had, at the time, already transformed the ability to search for timeframes, titles, and words in publications, newspapers, and archival papers – had advanced to a new level by enabling computers to read handwriting [3]. The development, known as Handwritten Text Recognition (HTR) software, is a subset of OCR and has proven invaluable for decrypting the multifarious and sometimes unclear penmanship of the past, such as clerks’ copies of wills, with their characteristic curls and flows.

The technology piloted by The National Archives in 2018 – which showed promising outcomes and significant potential, has expanded substantially. Named Transkribus, the system now collaborates with over 150 members spanning 30 countries – including researchers, institutes, archives, and universities – and has greatly upgraded the accessibility of archival records.

Traditionally, non-destructive professional book scanners are brought in to preserve fragile collections and extract details for digitisation. However, brands like Transkribus are pioneering methods to achieve similar achievements using smartphones and ‘tent’ configurations. 

In book scanners, a variety of optics are usually employed to precisely acquire information for OCR programmes to decipher. 

a set of colour filters sold by Knight Optical

They can i

nclude:

  • Polarisers: To reduce reflections and glare from glossy pages
  • Infrared (IR) Filters: To block or pass IR light to enhance certain features or lower noise
  • Colour Filters: To capture different colour channels separately for precise colour reproduction.

Advancing Healthcare Competence: OCR for Medical Records

Facts and figures written by clerks, scribes, academics, and writers of yesteryear aren’t the only intel that can be processed by OCR. Today’s healthcare professionals constantly take vital annotations on patients’ health statuses, and in some cases, such as pathology reports, this data is considered “difficult to mine.” Here’s where artificial intelligence (AI) steps in to help.

a doctor writing on a notepadMuch like the challenges faced by The National Archives in capturing manuscripts, handwritten pathologist notes, often scanned into PDF formats, are illegible to machines. But, for Nicholas Tatonetti, PhD, Vice Chair of Operations in the Department of Computational Biomedicine, and his team at the LA-based academic medical centre Cedars-Sinai, this obstacle presented an opportunity. Leveraging AI, Cedars-Sinai and The Cancer Genome Atlas were able to ‘tidy up’ scans, and OCR was then recruited to turn them into machine-readable notes. According to Tatonetti, the highly accurate conclusions “can help investigators identify and validate new disease markers, conduct research, and recruit patients for clinical trials [4].”

AI-powered OCR is not only saving clinicians time by automating and speeding up administrative tasks and, therefore, reducing the need for manual labour and associated costs, but it also enables quick access to digitised medical records, making it easier for medical professionals to retrieve patient histories, lab results, and other critical data during consultations.

OCR in Industrial and Machine Vision Applications

AI-charged OCR has extended its application across various sectors, amplifying productivity by accurately deciphering challenging text and characters in complex scenarios, including reading wording on curved or glossy surfaces, as well as embossed or engraved details. In manufacturing settings, OCR serves as a key tool, seamlessly integrating with largely automated assembly lines worldwide to streamline operations.

 

a render of a car assembly line utilising AI technology

In these surroundings, the tech combined with deep-learning OCR are enhancing production lines by increasing efficiency. The arrangements leverage high-resolution cameras and industrial scanners, adept at handling distinct surfaces, and utilise machine vision to swiftly capture and interpret text on products, packaging or elements in real-time. With the addition of deep learning, such as convolutional neural networks that mimic human visual processing, the systems gather and process data faster, augmenting operational speed and accuracy.

Optical Components for OCR Hardware

To ensure optimal performance of OCR tools, the hardware and its integrated components must be carefully chosen. Factors ranging from lighting conditions to surface properties significantly influence the effectiveness of cameras in acquiring information. For example, optics – such as Filters and Lenses – can be enhanced with anti-reflective (AR) coatings, particularly beneficial in industrial situations where shiny elements – such as automotive parts – frequently challenge cameras and vision systems. By incorporating AR-coated optics, cameras can recognise clear specifics, facilitating exact text recognition with OCR software to streamline processes.

a selection of optical components made by Knight Optical

Here at Knight Optical, we provide an array of coating options tailored to meet the precise requirements of machine vision systems and other optical applications. Our extensive portfolio includes substrates covering every spectrum – from ultraviolet (UV) through to infrared (IR) – each selected to excel in specific environments. For unique needs not met by our in-stock catalogue of over 3000 optics, we also offer custom-made optical components, ensuring seamless integration with your end products.

FOOTNOTES: 

[1] https://www.whatech.com/og/markets-research/energy/850553-optical-character-recognition-ocr-market-2024-2031-emerging-trends-growth-opportunities-growth-and-business-strategies 

[2] https://straitsresearch.com/report/optical-character-recognition-market#:~:text=Market%20Overview,USD%2012%2C567%20million%20in%202023

[3] https://blog.nationalarchives.gov.uk/machines-reading-the-archive-handwritten-text-recognition-software/ 

[4] https://www.cedars-sinai.org/newsroom/new-ai-tool-mines-cancer-patients-pathology-data/