Ensuring safe jam-making for delicious results

It is said that the average number of meals we consume during a lifetime is 90,000. This number is calculated on the assumption that we eat three times a day until the age of 80. Eating doesn't just support our lives, but also brings more color into life through the joy of eating.
For those reasons, every meal is very important. It is natural that we want to consume food that's as delicious and safe as possible, isn't it?

Awareness of the vital importance of food safety and security is increasing worldwide. Without exception, efforts throughout the food industry are concentrated on quality control using various methods. Given these circumstances, Nikon has developed a foreign material inspection system that utilizes optical technology and AI, providing unprecedented, innovative quality control solutions that help ensure safe jam making.

Distinguishing foreign materials using spectroscopic technology

Of course, in the food industry, careful measures have already been taken to ensure the safety of materials by preventing foreign materials during manufacturing.
Using metal detectors and X-ray inspection machines, metal fragments and plastic fragments can be detected. However, detection of organic substances (substances derived from plants and animals) mostly depends on human eyes. This brought Nikon to focus on spectroscopic technology.

Spectroscopic technology enables measurement of the spectrum of reflected light from an object (distribution of the intensity of light wavelength components), then applies that information to various fields. We recognize the light of the wavelength that an object reflects as "color". For example, strawberries look red because they reflect a lot of red light (long-wavelength light) and absorb other light (short- to medium-wavelength light). Which wavelength is reflected and which wavelength is absorbed (spectral reflectivity characteristics) differs depending on the characteristics of the object.

Utilizing deep learning to improve inspection accuracy

In jam and fruit spread manufacturing, there is a process to eliminate foreign objects and impurities contained in the materials. Until now, that inspection has been conducted by human eyes, however, there were several issues such as the heavy physical burden on employees and inconsistent detection accuracy. Due to the wide variety of raw materials and differing shapes of fruits, it was considered extremely difficult to automate this inspection process.

Supporting customers, and contributing to food safety

I led the development project as a team leader. In 2015, we received a request from AOHATA Corporation about the feasibility of an automatic inspection system using Nikon's spectroscopic technique. After discussions within the team, the prospect of proceeding with the project using spectroscopy technology seemed within reach, so we started to study the specifications from 2016. After repeated experiments, we brought the first prototype of the equipment into the actual factory, however the environment was very different from that in the laboratory, and we could not get the detection accuracy we wanted. Therefore, the image recognition method by AI was changed from machine learning where a person defines parameters to deep learning in which AI learns by itself, and the detection rate then dramatically improved with a reduction in false detection. It had successfully reached the level of practical use.

I am a mechanical designer, so I always enjoy knowing that my design has achieved the form of a product, and becomes something tangible. And when customers are happy, I couldn't be happier. I would like to continue to develop such products and systems into the future.

This foreign material inspection system was the first product I was involved in, and I was in charge of electronic design and software. There were two things that I found difficult during development.
One was the flash light source. It was not easy to find a light source that has excellent light intensity, uniformity and durability. Cooperating with light source manufacturers, we carefully conducted testing and analysis. I felt very relieved when I finally found the solution.

The other was the problem, on very rare occasions, of the software processing becoming heavy when used for a long duration, which makes it impossible to continue shooting. That was my first experience, so I was very nervous about it. Carefully listening to the situation from customers after test installation, I analyzed logs, optimized and then was able to stabilize the situation.

For the development, since I was involved in consecutive processes from specification development, design, follow-up with the manufacturing department, and adjustment at the manufacturing site, I felt a great sense of accomplishment. I would like to create products that can be a pillar of Nikon from this business unit.

As a sales representative, I was a contact person with AOHATA Corporation. To realize their requests and demands, I supported the project in a wide range of fields both within the office and with AOHATA's partner companies including participating in meetings about the specifications, managing schedules and attending the delivery of the components.

AOHATA Corporation is extremely particular about food quality and values its safety and security very highly. I kept this constantly in my mind to understand their attitude and managed all the details.

Thanks to AOHATA Corporation, as well as members of my office, and the knowledge and cooperation from people of other business units, this equipment was successfully realized.
I intend to continue utilizing Nikon's internal technologies and trying to cooperate cross-organizationally. I would like to offer products that make everyone happy.

Originally published: June 28, 2019.