FROM CHALLENGE TO DELIVERY
OCR Installation for Labels and Shop-Floor Data
An OCR system reads characters from images and sends the results on for further use. We design the camera, lighting, and software to suit your labels or reading points, testing with real images and agreeing on how to handle unreadable cases before installation.
Suitable for: Suited to work that needs to read labels or data from images in order to record them or compare them with an existing system. Work that does not yet have sample images or acceptance criteria should start with testing before specifying the system to be installed.
Work Illustration
From reel labels to data for barcode printing
We developed a system that reads reel labels, checks them against master CSV data, and prints barcode stickers. At the start of use, Manual mode worked normally, while Auto mode needed further reading adjustments based on customer feedback. Prepare sample labels and agree on how to review the readings before starting. The Edge AI meter-reading case remains a prototype, not an example of a system installed for actual use.
Our past projects
What you receive
- Scope of the characters and data to be read, together with sample test images
- Camera, lighting, and imaging position design based on site conditions
- Data reading software and integration with the target system as agreed
- Acceptance criteria, procedure for verifying read results, and user manual
Suitable for: Suited to work that needs to read labels or data from images in order to record them or compare them with an existing system. Work that does not yet have sample images or acceptance criteria should start with testing before specifying the system to be installed.
Our process
Define the data to be read
Prepare labels or sample images, both the easy-to-read ones and the problematic ones, and specify the target system and the speed required on site.
Test the camera and the reading
Test lighting, camera angle, and OCR results with real samples, then agree on acceptance criteria and how to handle misreads.
Integrate and hand over
Link the read results with existing data, test together with the users, and hand over within the agreed scope.
Frequently Asked Questions
Can OCR read characters on reflective surfaces?
The camera, lighting, and shooting angle must first be trialed on the actual surface. Reflections and unclear characters may make reading impossible, so accuracy cannot be specified from a job description alone.
What kind of images should be prepared to assess the work?
Prepare sample images from the actual usage position, including different label types, small characters, reflections, and hard-to-read cases, and specify the data fields to be read along with examples of the correct results.
How do OCR results enter the existing system?
The data format and the integration channel are defined together with the team that maintains the existing system. In the reel label case, the system compared the read results with a master data CSV file before printing stickers. Other work requires the integration channel to be assessed separately for each system.
How are misreads evaluated?
We jointly define a set of test images and the correct results, then check the cases read correctly, read incorrectly, and not readable, and agree on a procedure for users to verify results before the data is used further.