This deep learning label inspection application is an automated machine vision system for printed labels to eliminate human inspection, improve quality, and production efficiency. The system performs automated Deep Learning based OCR/OCV, 1D/2D code reading and Print Quality Inspection (PQI).

Operator can train labels of varying sizes and printed text/graphics to generate their own recipe through an easy-to-use training User Interface.

Features:

• The system utilizes high resolution linescan imaging to inspect pharmaceutical labels with high accuracy.
• System implements a Deep Neural Network (DNN) to perform OCR on text as small as 5pt in size. The OCR/OCV is agnostic to font size/style, character spacing, character size within the same string. There is no need to train any characters, making the system extremely easy to use.
• System reads and grades designated code 1D and 2D symbology at various regions on the labels.
• Detects blemishes in print (text and graphics) in designated areas of the label where the defect is present.
• The system uses PDF print files for each label as the “first article” template for that label. The PDF file is used as a golden template to define the regions of interest for various inspections to be performed on the label: OCR/V, blemish etc.). A recipe (set of label specific parameters for each label type, or label configuration) is trained based on the PDF and loaded up before running the roll.
• The system fully handles the roll printing and inspection. Blank label stock is loaded on, good, printed labels are dispensed, and rejected labels are rolled into the take-up roll.

View machine vision inspection systems from Integro here or on YouTube.