HEUFT reflexx A.I. X-ray image proessiing

HEUFT reflexx ^A.I. is now even smarter! A newly integrated deep learning algorithm goes into greater depth during the X-ray image analysis – and makes dangerous foreign objects detectable even where they could not previously be recognized.

Artificial intelligence (AI) has been used in HEUFT inspection systems for over ten years to ensure reliable detection, smart multidimensional evaluation and reliable classification of a wide variety of objects in the food filling and packaging process. In the case of disordered structured products such as gherkins or red cabbage in food jars, machine learning is not always sufficient to identify dangerous foreign objects such as glass splinters in particular and distinguish them from harmless product components and inhomogeneities.

The latest version of the in-house developed hardware and software for intelligent X-ray image processing, which is now available for foreign object detectors in the current HEUFT eXaminer ^II series, changes this: with a new deep learning algorithm, HEUFT reflexx ^A.I. makes previously invisible objects visible even in places where this was previously not possible: in inhomogeneous product masses with irregular structures and cavities between their individual components with different levels of absorption.

With the help of the new deep learning algorithm, the aluminium fragment between the gherkins can be detected for the first time just as reliably as the stone in the red cabbage or the ring-shaped piece of cable in ring-shaped pasta of the same size. HEUFT reflexx ^A.I. finds and marks the dangerous foreign objects in real time – and at the same time reliably distinguishes them from harmless product and packaging structures so that the false rejection rate during the pulsed X-ray inspection tends towards zero.

HEUFT has combined tried-and-tested image analysis and AI processes with a multi-layered neural network that goes into greater depth and thus processes even abstract patterns independently in a meaningful way. This makes the deep-learning-capable HEUFT reflexx ^A.I. far superior to conventional analysis methods such as gray scale determination, contrast detection and machine learning for the detection and classification of different objects.

Previously invisible objects become visible even under difficult environmental conditions – and the proportion of false rejections of actually uncontaminated products is reduced; valuable packaging or food is no longer wasted in vain.