Product

Down with false positives!

14 Feb 2018

Marel’s latest SensorX software

For poultry processors, the ideal X-ray detection machine should detect bone remnants and other contaminants with the very highest degree of accuracy. What processors don’t want is a machine which gives too many false positives (a bone-free piece is wrongly rejected).
 

SensorX bone detection system

SensorX Poultry
 

Marel’s new SensorX software creates the perfect balance between highly accurate bone detection and the lowest rate of false positive.

With too many false positives, bone performance detection statistics – while looking impressive – will not reflect the true situation and could be misleading for departmental management. Marel’s latest SensorX software not only reduces the incidence of false positives to an absolute industry minimum but also detects those difficult to find bone fragments more accurately.

This means less rework and fewer inspectors, enhancing the machine’s already excellent performance.
 

PERFECT BALANCE

The major advancement in the new SensorX software is that it intelligently combines algorithms with new automatic calibration and a very accurate X-ray sensor, in order to get the most reliable detection.

The new software enables the SensorX to create a perfect balance between an increased detection rate and a reduced false positive rate.

However, this balance is a very thin line. When trying to detect the very smallest bone particles, a machine runs the risk of generating erroneous results and jeopardizes its reliability.

Therefore, the first aim of Marel was to detect the overall highest percentage of bone particles – big as well as small – in the market.

Elaborate testing of the new SensorX software shows that this objective has now become a reality.

When performing bone detection tests, false positives should always be included. If detection results are excellent but the false positive rate is 30%, the machine is useless, because there’s way too much unnecessary rework to be done.
 

SAVING LABOR

Bone detection

A dual lane SensorX, working at full capacity, handles 240 pieces of meat per minute. Let’s assume the bone detection rate is 5% and the false positive rate is 6%. That would mean that 26 pieces of meat are rejected per minute. As an average operator can inspect 10 pieces per minute for bones, three of them would be needed to keep production running.

When you succeed in bringing down the false positive rate to 3%, this would result in 19 pieces per minute, needing only two inspectors. In this way the labor saving effects of the new SensorX software are immediately visible.

Even further downstream in the process, right up to the processor’s customers, the SensorX saves labor.

Brad Knipfel, Operational Manager at Cargill London Ontario, Canada, says, “Customers who in the past have received both x-rayed and non x-rayed product from us no longer want the latter. X-rayed products saves them labor.”

Continuous improvements in the form of still more new, intelligent software applications will keep strengthening the position of Marel’s SensorX as market leader.