Support for sandwich sorting

Share this on social media:

Stemmer Imaging has successfully tackled a challenge to automate a process to check that sandwiches packaged for sale in a supermarket corresponded to their labels. The company undertook a feasibility study, which required that the sandwich was imaged through the 'clear' plastic packaging, while the label on the packaging was also read.

Reproducible identification of sandwich contents and bread types was achieved using the Manto pattern recognition tool, which is part of Stemmer Imaging's Common Vision Blox hardware independent machine vision toolkit. Manto is an object recognition tool that uses a neural technology resulting from research in the field of artificial intelligence, especially statistical learning theory.

Manto differs from other pattern recognition approaches, because it allows, among other things, the recognition of organic forms and textures that show a high level of distortion or that are hard to reproduce. This makes it particularly well suited to this application as it copes well with applications where other methods of pattern recognition fail.

The system is trained to identify a particular sandwich type by presenting it with a series of training images of the sandwiches, until the system achieves acceptable levels of sandwich recognition. Unlike traditional neural network technology, Manto has the ability to use any number of training images, thus providing the ability to classify objects
with an accuracy not previously possible.

In the feasibility study, the system was successfully trained to recognise cheese Ploughman’s, cheese and celery, and cheese and bacon club sandwiches.