Machine Vision System Scans Baked Goods

Daniel S. Burgess

Engineers at Georgia Institute of Technology in Atlanta are working on a way to give everyone perfect buns -- sandwich buns, that is. In collaboration with researchers at Baking Technology Systems Inc. (Bake-Tech) of Tucker, Ga., they are developing a machine vision system that continuously inspects fresh-baked buns to ensure the proper and uniform color, shape and seed distribution.

A machine vision system inspects sandwich buns on the production line at Flowers Bakery in Villa Rica, Ga. Courtesy of Georgia Institute of Technology.

Doug Britton, a research engineer at the institute and the co-principal investigator on the project, explained that customers of the baking industry desire an automated means to verify product compliance with quality standards. Currently, baking operations grade their goods on the production line by manually inspecting a small sample using handheld colorimeters. An automated machine vision system, in contrast, would avoid this labor-intensive process and would inspect all products exiting the oven, he said. Moreover, it would offer a source of adaptive feedback data that could be used with proof-box and oven controllers to increase yields.

The vision system, which is undergoing testing at Flowers Bakery in Villa Rica, Ga., comprises two Dragonfly 1/3-in. progressive-scan CCD cameras from Point Grey Research Inc. of Vancouver, British Columbia, Canada, a PC running Windows 2000 as the image-processing unit and another PC for database storage. The researchers developed the image-processing software in-house specifically for the application.

Britton said that this system offers three advantages to users. It can be placed on the cooling line in the baking process, which reduces infrastructure costs by eliminating the need for a dedicated, specialized conveyor belt for inspection. Second, it employs low-cost color cameras connected by FireWire interface to the image-processing unit, easing networking and camera synchronization. Lastly, it stores the collected data in SQL, a standard database language, making it possible for approved users on the network to generate production reports -- such as the number of baked goods produced or the number that failed to meet quality standards -- over an Ethernet connection.

Refinements

Future development will apply this data to enable the system to synch with proof-boxes and ovens for inline process control, automatically adjusting the rising and baking conditions to minimize product flaws.

The researchers are working on a hardened version of the system for long-term testing on a bakery line to give them feedback on how to improve the user interface and the report generators. Such a system should be suitable for commercialization and for use with any baked goods that are transported on conveyor belts, Britton said. The researchers also are exploring potential applications beyond baking, including in the production of value-added cooked and partially fried goods for the meat and poultry industries.