Spectrim Automates Quality Control with Machine Learning
Kagan Pittman posted on January 22, 2016 |

The Spectrim optical sorting platform recently launched for the food and beverage industry. The platform uses machine learning software to perform quality checks of fruit.

The Spectrim optical sorting platform. (Image courtesy Compac.)
The Spectrim optical sorting platform. (Image courtesy Compac.)

Spectrim can detect hard to find external blemishes on fruit and automatically separate produce by grade using existing technology.

The platform is capable of taking up to 500 high definition images of a single piece of fruit as it passes through the sorting machine at a rate of 12 pieces of fruit per second.

Similar machine learning technology has been used before by Google’s DeepDream software and Bonirob’s farming robots.

Blemishes, bruises and rot on apples are highlighted and classified. (Image courtesy Compac.)

Blemishes, bruises and rot on apples are highlighted and classified. (Image courtesy Compac.)

Operators must first teach the Spectrim platform the difference between good and bad apple skins. Spectrim can learn to identify blemishes after seeing examples highlighted with a click and drag interface on an already scanned fruit.

Spectrim memorizes the appearance of defects and can even be used to classify different defect types. With this capability, operators can order Spectrim to create sophisticated pack grades to separate incoming fruit according to quality.

Highly detailed scanning within the sorting machine is possible thanks to uniform lighting and high quality optics. To process images, Spectrim uses multiple wavelengths, which can be configured to target specific defects.

Compac, a producer of packhouse technology for the produce industry, designed Spectrim for integration within their own existing products. New Zealand’s Horticultural Crown Research program offered critical input for the creation of the platform.

What Other Industries Utilize Automated Quality Control?

Spectrim’s technology is not out of reach for large automation providers like FANUC, Mazak and KUKA Robotics. So why hasn’t this kind of technology expanded outside the food and beverage industry?

The thing is, it has – just not with this level of machine intelligence.

For the automotive industry, GOM offers automated quality control in the form of Virtual Measuring Room software. With this software, a robot arm can scan the dimensions of a vehicle part using parametric inspection software.


GOM also offers a solution for inspection of small parts with their ATOS industrial 3D scanning technology. ATOS 3D scanners will asses a component’s entire surface geometry for operators to distinguish warpage and defects.

What sets Spectrim apart from these automated solutions, however, is the apple-sorting platform’s machine learning capabilities.

What if ATOS technology could perform complete inspections, separating good parts from bad without the need of a human analyzing scan results?

With maybe just a couple decades worth of advancement, manufacturing facilities could make use of this machine learning technology to the point of fully automating facilities. A human operator’s only job at that point would be to ensure the software is working as intended.

For more information on the Spectrim optical sorting platform, visit www.compacsort.com/spectrim

Recommended For You