The 3-D way to slice it

Jörg Schwartz, joerg.schwartz@photonics.com

They may all look the same in their transparent packets in your supermarket’s refrigerated cases, but hams, sausages and cheeses are nowhere near uniform, and slicing them remains a challenge for the food industry. Today’s fixed-volume or fixed-weight approaches deal with shape variations and require significant margins, factors that cause more food to be placed in a pack than is needed. That extra cost to the manufacturer eventually is borne by the customer.

A new three-dimensional imaging system developed by Spanish 3-D software company AQSense and Italian solution provider ImagingLab srl uses three line-laser emitters and three cameras to generate a contour of the whole object before it is sliced, with less than 1 percent measurement error. Demonstrated at Vision Show 2009 in Stuttgart, Germany, in November, it calculates the best way of slicing while the job is being processed, at speeds in the range of 10 m per minute.

Three-dimensional measurements are a hot topic in machine vision (see “Vision Market on the Upswing” at photonics.com), with quality control a key area of application. Industrial 3-D measurement systems typically generate “point clouds” of data, with the size of the cloud depending upon the object and the resolution of the measurement system. The bigger the cloud, the more demanding the processing.

Typically, the goal of such data handling is to compare the test object with a computer-aided design model to see whether it has any defects, a process that determines a pass/fail decision. Doing this at high speed for complex objects is demanding, but rapid progress is being made on the hardware side, with faster and more intelligent cameras and computers.

New algorithms also are enabling quick matching of 3-D objects, somewhat similar to 2-D pattern recognition. “Our goal is to offer a comprehensive library such as those available for 2-D machine vision,” said Ramon Palli, AQSense’s CEO.

He said the challenge is not only in calculating the full shape, but also in extracting the data from the measurement. “Laser triangulation is a most widely used detection method, but for many surface types, the data is not directly usable and requires calculating the peak in the reflected laser light before that data can be used.”

A laser-triangulation-based three-dimensional scanner generates a point cloud of measurement points, used to determine the shape of cheese or ham before it is sliced. This example shows the growing capabilities of 3-D imaging, while making food preparation and processing more efficient.

Such is the case for surfaces where the light is not just reflected but where it partially penetrates the material; e.g., when the reflected light has a non-Gaussian profile, and the peak intensity and center of gravity are not the same.

This know-how and that of 3-D shape analysis – in the form of the company’s software library SAL3D – were combined with the hardware and system integration skills of ImagingLab when developing the 360° scanner prototype for food processing. ImagingLab also wrote LabView application software, including sample handling and user interface.

Both companies say that slicing food is just the beginning. The combination of 3-D machine vision and robotics is needed for flexible manufacturing in a wide range of areas – with both their 3-D library and the application software being compatible with any other 3-D imaging device generating point clouds.