The Latest Trend In X-Ray Inspection Systems: The Evolution Of AI-Driven X-Ray Inspection Technology

Tuesday, March 5th, 2019

Introducing changes and trends of X-ray inspection systems that continue to evolve while incorporating new technologies. By Ishida.


The demand from retailers and consumers for manufacturers to ensure food safety is increasing like never before. To meet this demand, manufacturers can now rely on X-ray machines to detect objects invisible to the naked eye due to the development of detection technology.


The Definition Of “Foreign Object”

The term “foreign object” generally refers to any physical object from outside which has contaminated a food product. In some countries, they are defined according to their size. For instance, the American Food and Drug Administration (FDA) Compliance Policy Guides contain the following statement: “The Board found that foreign objects that are less than 7 mm, maximum dimension, rarely cause trauma or serious injury except in special risk groups such as infants, surgery patients, and the elderly. The scientific and clinical literature supports this conclusion”.

In Japan, 1000 food products are recalled every year. Fifty percent of these are for products that have been printed with the wrong best-before date or the wrong ingredients, while 15 percent is due to foreign object contamination. The contamination of food by foreign objects, which can be harmful to consumers, can be damaging to a brand’s credibility, so inspection for foreign objects at the production stage is becoming an essential part of the process.


Changes In Foreign-Object Detection

In the food manufacturing industry, inspection systems—mostly metal detectors—have been used to inspect products for foreign objects since the 1990s. The main aim at the time was to detect foreign objects that had contaminated the food due to wear or damage to the factory equipment. Around the year 2000, due to a rise in consumer awareness of food safety, there was a rising demand for detection of non-metallic foreign objects contamination in food at the site of production, such as small stones or pieces of glass in agricultural products. Stakeholders became increasingly interested in X-ray devices that can detect foreign objects made of materials other than metal.


There was also a growing need to detect organic matter, such as hard plastics and rubber, and unwanted food-borne objects found naturally in food—chicken bones, for example—but no existing X-ray inspection systems were capable of detecting such things. This set the stage for the development of dual energy technology: a method in which two types of image are captured, and the foreign object is detected by comparing the properties of the images. This epoch-making technology achieved higher inspection accuracy on bones which were difficult to be detected before.

In recent years, there has been a demand for controls that are capable of removing even lower-density materials, such as cartilage, hair and insects, but the reality is that these objects are difficult to detect, even using dual energy technology.


The Evolution Of Image-Processing Technology

Dual energy did not just improve machine sensing technology—it was a revolution in the field of foreign object detection. On the software front, another important element of X-ray foreign-object detection was continuing to develop: image processing software. The first X-ray devices used ordinary image processing methods, but in 2004, the Ishida IX Series was released, equipped with cutting-edge genetic algorithm technology. These technologies made it capable using computers to generate the optimal image-processing methods for differentiating between food products and foreign objects automatically, and apply the most appropriate type of image-processing on a case-by-case basis. Now, food could be inspected with a high degree of accuracy. Over the years, X-ray inspection systems have begun to employ machine learning, and became able to estimate the product weight. They can use this data not only to inspect goods for foreign objects, but also to check for missing items and to sort goods by grade. These technologies can broadly be called Artificial Intelligence (AI).


The Application Of AI

Products that make use of AI have been in development for the last 15 years, and the recent leaps forward in the field of deep learning have raised questions about how this technology could be applied to the detection of foreign objects. In fact, an X-ray inspection system incorporating AI was presented by Ishida at an exhibition in 2018. In general deep learning, a large number of images are labelled in advance by a human user (in foreign-object detection, this means pointing out the location of the object manually), and then used to train the computer. This learning process requires highly complex computing, but deep learning operations resemble the ones used in computer graphics and because of the high affinity with the Graphic Processing Unit (GPU), it is looking increasingly likely that the rapid development of information technology in recent years made AI learning a reality. The Ishida team is still at the testing phase, but we have come to understand that GPU-based deep learning can be used to carry out advanced foreign-object detection operations with the IX Series, just as other systems such as internet search engines and automatic driving systems acquire sophisticated capabilities. By applying AI, X-ray inspection system will be able to detect even extremely difficult-to-detect foreign objects, including certain types of bone that manufacturers have given up on finding.

When using deep learning, only the foreign object itself is detected; the areas where the sausages overlap, which are often misinterpreted by existing technology, are not detected. The image of the chicken breast illustrates how foreign objects located in darker areas of the image are picked up.


Possibilities Of Application

AI is developing rapidly, and is expected to continue to do so, but the technology remains elusive. One reason for this is that it is almost impossible for humans to understand the algorithms involved with the test results performed by AI. The technology that handles this is also becoming more and more advanced, requiring specialist knowledge. On the other hand, AI is certainly an effective means of developing more precise X-ray inspection technology, and there is a strong demand for the application of AI in the food manufacturing industry. Inspection-machine manufacturers will surely accelerate the development process along.


On the hardware front, it is hoped that dual energy technology will be further developed. If high-definition scanning and advanced energy analysis become available, it will become possible to detect objects that were previously undetectable. Foreign-object detection is expected to continue evolving through the development of software such as image processing, AI and the improvement of hardware such as sensing technology.


Foreign-object detection continues to be essential for safeguarding the quality of products in the food manufacturing industry, and we believe that the means of detection employed will continue to diversify, and will go on to be used even more widely across the world.