Why the Anyline SDK will revolutionize Mobile OCR
For a long time it has been “online vs. offline” and “digital vs. analog” – but we are in the midst of the all-digital age, part of a generation which strives to turn towards a mobile device and the internet when in search of anything. There have been a lot of different approaches towards finding the perfect interface between the world around us and the objects we can touch and the connected world. “Internet of Things” has become the buzzword of our time as we try to invent the universal interface. There have been many partially working options: voice recognition to transfer written information into digital data, bluetooth for transferring data from one small device to another and many others. Text recognition has been around for a long time, but Mobile OCR (Optical Character Recognition) did not have the expected capabilities so far.
Today we are proposing the most generic, but also in Anyline’s case most accurate new interface between the two worlds. Anyline® uses mobile cameras to capture written text, data and numbers and converts them into digital output in real-time. An advanced and highly robust mobile OCR technology which therefore only needs one reliable interface: the camera.
What is Mobile OCR?
Mobile Optical Character Recognition describes an OCR process which is entirely executed on a mobile device. All image pre-processing and text recognition algorithms work without a server connection and only use the mobile device’s camera solution and processor power.
OCR exists since 1980s, Anyline adds accuracy
Optical Character Recognition has truly been around for two decades already, this is common knowledge and we like to acknowledge the fact (for all the nay-sayers ;)), but it has never been put to work in highly critical circumstances. Anyline® has not been designed to simply read text with a moderate detection rate to give back an approximate value, but to dive into all those real-world use cases where accuracy is a key requirement for OCR to be used.
Anyline basically helps Mobile OCR in growing up. We are gradually moving away from reading data within standardized and controlled circumstances, towards reading information with a tool we always have at hand: the smartphone or wearables like smart glasses.
The innovation of the patented Anyline technology lies in its advanced image pre-processing. Image pre-processing describes everything that happens before text recognition and OCR algorithms come in:
– Detecting a three dimensional object and the cutout with the desired data in it,
– Adjusting the perspective, correcting the distortion, obliterating reflections and many more,
– Standardizing the video frames within real-time and
– Delivering the frames ready to be read with the identified text easily recognizable in black on white to the character recognition algorithms.
This was the biggest technical obstacle in our research and development, as the processing has to happen very quickly due to constantly changing lighting conditions, and because of the different qualities of smartphone cameras. Still, nowadays we’ve even tricked the auto-focus of about 50 Android devices and can proudly say that we can officially support them if needed.
This is why Mobile OCR has not been taken seriously until now – it has always worked under completely controlled conditions, but not on a mobile devices, where situations and technical specifications can not be fully predicted.
Process optimisation and quality assurance use cases with Mobile OCR
There are common use cases like the very well-known “WordLens” which has been acquired by Google to translate text or business card scanners on mobile phones which use great OCR algorithms, but we have focused on industrial and medical applications.
Our first project with mySugr, a company which provides apps for people with diabetes has taught us a lot. For them we’ve developed the mySugr Importer – an application where Anyline detects the glucose level, date and time on a lot of different kinds of blood sugar meters and imports them when filming the display. In this very case we have to be sure, we make no mistakes – therefore the recognition process takes a few seconds, but with this kind of sensitive medical data, accuracy is more important than speed when digitalising a value.
Furthermore we are looking at the industrial sector for applications. Here Anyline can add significant value in optimising processes, by substituting manual actions like writing down serial numbers and putting them in in cumbersome ways. This means, when recognising barcodes, serial numbers, type numbers and all kinds of written data on assembly parts or tools, Anyline can be crucial to creating context for a worker in the field and automatically help out with a decision support system in the back.
The Revolution is coming
Not only are off-the-shelf smartphones and wearables getting better and better, we are also used to carrying them with us as a tool all the time. The acceptance rate of a smartphone which adds significant value to everyday life or a wearable to helping out in work situations is advancing and plays right into Anyline’s plan to take over the world (of Mobile OCR).
QUESTIONS? LET US KNOW!