In the months since we first announced the arrival of the incredible new Cognex In-Sight® D900, the true power of this new camera has continued to amaze as we’ve rolled out the first trials and customer projects.
The sheer range of inspections that it’s capable of within the standalone camera is incredible – all thanks to the built-in ViDi™ deep learning software that takes machine vision beyond human and rules-based image inspection.
We’re talking about carrying out assembly verification, defect detection and optical character recognition; learning and making decisions with ‘human-like’ ability in a way that’s never before been so commercially and financially available.
It has to be seen to be believed.
Request a demo[hubspot portal=”9430593″ id=”680cb951-1a32-489e-9fa0-88b5c7e38664″ type=”form”]
Back to basics: what is OCR automation and why does it matter?
The labelling and coding of products have become essential within the supply chain. It allows facilities to use warehouse management systems to track and inspect shipments, parts and products. Before automation, handwritten and manual systems were time-consuming and at high risk of human error.
Automated label readers have significantly increased processing speeds and accuracy with the ability to read and log data within seconds. The automation of label reading has become possible due to the development of vision automation. Camera systems are now specifically being designed for use within production and fulfilment centres.
For simple labels that are easy to read, low-level specification cameras and setups are sufficient. Some facilities just need to know the order number and where the item should be placed or dispatched.
However, what happens when the label or text is difficult to read? Advanced technology is needed for difficult and challenging optical character recognition (OCR) such as parts or poor quality labels. This need has led to the development and application of deep learning technology to extend the capabilities to difficult OCR tasks.
Deep learning-based vision software for OCR
Deep learning-based vision software allows facilities to easily and quickly solve challenging OCR, verification and defects detection. The intelligent software system allows facilities to set up rule-based machine vision tools for fast and reliable readings which are not possible using human eye inspection.
By showing the camera multiple examples of both good and bad standards, users can teach automated camera readers what characters, text and labels should look like to detect defects. This allows for the automated diversion of faulty or incorrect items for improved processing accuracy.
Deep learning technology allows the camera to read difficult labels including those with poorly marked characters or low contrast characters to avoid incorrect dismissal.
The Cognex OCR In-Sight® D900
Powered by In-Sight ViDi deep learning software, the D900 is a vision system designed to run state-of-the-art algorithms for industrial image analysis. There are two versions of the D900 available, a lower-cost OCR version for reading without any recognition challenges and a second advanced model with additional tools.
The advanced D900 model has all the standard functionality of the Cognex In-Sight smart camera plus the ViDi deep learning software. The powerful combination allows for smart capabilities including taking measurements, detecting edges, reading barcodes, blob inspection and pattern finding.
The deep learning technology is designed for ease of use and operators do not need extensive knowledge to gain the most out of the camera systems. With user-friendly interfaces and minimal initial programming or setup, the cameras are quick and simple to integrate.
Both versions of the D900 utilise a web-based HMI system which removes the need for any dedicated software to visualise and analyse images and results. It can be accessed by any PC for ease of accessibility.
Applying deep learning capabilities to processes
There is an increased demand in the automotive industry for advanced vision system technology. Car parts usually have part numbers printed on them which can be small and difficult to read with the human eye. The deep learning capabilities of the D900 allows it to easily and quickly recognise and read car part codes from pre-taught examples – even with the most difficult fonts.
Deep learning and vision automation are also highly beneficial to production processes and materials handling. Particularly popular on bottling lines, the D900 is capable of reaching processing units of 35,000 units per hour.
Continuing with production, canned drinks have an expiration date and lot code printed onto the bottom of them. Verifying that it has been printed correctly is an important job but also difficult and time-consuming. Using vision automation such as the D900 and applying deep learning allows the accurate and reliable checking of printed codes at high speeds. With the optional IP67 lens cover, the D900 can be used in wet environments making it ideal for food and drink production.
High-speed reliability and inspection
Overall, the D900 camera with a deep learning application offers higher accuracy as it’s able to read labels and codes varying in quality. Furthermore, because deep learning systems are pre-trained, you’re not required to teach them individual characters or fonts like traditional systems so it considerably reduces set-up time.
In addition to reading labels and codes, the D900 and ViDi deep learning system can also automate the 100 per cent inspection of products. This saves a significant amount of time compared to carrying out traditional manual inspections.
If you’re interested in finding out more about the Cognex In-Sight D900 vision system or understanding how it can assist with your production requirements, contact Bytronic.