Industrial bakery increases throughput and lowers errors with machine vision

(See Zebra Technologies at FoodManufacturing.Live, 1 May 2024, on stand 41)

The bakery can now inspect its full range of breads using a single machine vision solution, and carry out efficient, automated picking with a robotic grip handling between 25 and 30 packages per minute, without damaging bread or packaging. Its estimated that the new solution has secured a 75% cost saving compared to traditional camera and lighting inspection approaches.

The industrial bakery approached KINE, a Finnish headquartered provider of turnkey robotics solutions for the food and beverage, logistics, semiconductor, and other manufacturing industries. Together with OEM Finland Oy, a Zebra registered industrial automation system integrator and an advanced machine vision specialisation partner, they were tasked with developing a solution that would overcome several challenges.

A manual process was being used to remove and deposit bags of bread loaves and rolls, which was potentially error-prone and less efficient as irregular bread production sizes and shapes made it difficult to determine their location and orientation on the conveyor belt. Transparent plastic packaging also made detection difficult with optical sensors due to low contrast levels and partial reflections.

We decided to leverage Zebras Aurora Design Assistant machine vision software for its versatile, powerful capabilities that provide the robustness and integration needed to visually inspect the variations in packaged bread, said Kimmo Salonen, chief technology officer at KINE.

The new system consists of a PLC, robotic grip, 3D time-of-flight camera and Aurora Design Assistant software, which has eliminated the need for multiple cameras and lighting and reduced contrast and reflection problems.

Basic implementation of the vision system took about one day, said Sami Sinisalo, robotic specialist at KINE. Testing, adjusting, and fixing took about 80 hours, including time calibrating the robot. Thats a great turnaround and drastically less than what we'd typically have to do for a new 3D install without powerful and user-friendly software.

The camera is placed above the conveyor belt, with inspection flowcharts created in Aurora Design Assistant. The camera takes up to 30 frames per second, recording the bread surfaces as a point cloud with over 300,000 XYZ coordinates. The data is sent to Aurora Design Assistant which converts the data into a depth map for analysis, using 2D vision tools to determine grip points for the robot.

Challenging manufacturing environments with high levels of compliance and wide variations in components and finished products require powerful machine vision systems to deliver needed outcomes, said Jason MacDonald, EMEA senior account manager for machine vision at Zebra Technologies. Zebras high-performing Aurora machine vision software is moving industries forward and overcoming the problems of legacy systems faced by customers and machine builders.