December 27, 2021

The Electrical Science and Engineering Promotion Award is an honor presented by The Promotion Foundation for Electrical Science and Engineering. Since 1952, the foundation has continued to commend persons of merit who have contributed to Japanese electrical science and engineering. In the 69th Electrical Science and Engineering Promotion Award, all three of the entries from NEC were selected for awards, and the only one with top evaluation was specially selected for the Minister of Education, Culture, Sports, Science and Technology Award. In this article, we interview the researchers who receive the Minister of Education, Culture, Sports, Science and Technology Award about the thrill of receiving this special award.

Awarded technology

The Development and Commercialization of a Video Search Technology Discovering Unregistered Target People from Million-Scale Videos in 10s

The world's first technology achieves high-speed and high-accuracy searching of unregistered target people from a massive volume of video data. This technology uses biometric information to group the same people who appear multiple times in the videos of different locations.

Reaction from the award recipients

Shoji Nishimura

Joined NEC in 2001. Dr. Nishimura specializes in data engineering and databases, and has received numerous awards for research papers at conferences. At present, he has been transferred to the Planning Division where he is actively supporting the implementation of NEC research technologies in society. He has pursued research in the same team for close to ten years together with Dr. Jianquan Liu who shared this recent award.

Q1: So how do you feel about receiving this award?
I am extremely happy. I feel glad to have chosen the path of becoming a researcher. In particular, we have experienced frustration during the development of this technology, so I am deeply moved by receiving this award. The development of this research started in 2012. It was right around the dawn of the third AI boom. We speculated that high-speed video search would be needed in the future and got started. We quickly completed a prototype of this technology and were able to go so far as to commercialize it. However, customers at the time had not accumulated the data, and there was no demand for the technology. Therefore, the planned field testing for a proof of concept was unfortunately terminated. For about a year and a half after that, we reconsidered the concept and worked on investigating customer needs.

Q2: What do you think led to this year's award?
While experiencing a research setback, we came up with the idea of changing how we view the technology. Instead of being preoccupied with rapidly searching for targets from a large volume of data, we focused on the "functionality of grouping similar objects" possessed by this technology. The technology was the same, but we changed how we approached it. I think that this was the turning point. When similar objects are grouped together, you can detect people appearing at the same location and know the "frequency." From there, you can tie that to behavior recognition such as loitering and lost children. When it was presented in this way, people from the business units became very interested, and even the executives said that it was an amazing technology. Furthermore, for a team such as ours that is involved in and doing the database research at the backend, we also created a compelling example of how we can take center stage.

Q3: What are your goals going forward?
I has been transferred to the Research Planning Division where I am currently in a position to oversee technology for all of the research laboratories. I hope to utilize my experience cultivated in technology research to promote the "redefinition of technologies from a different perspective" in other research areas as well. There are many technologies in the research laboratories, but I hope to help make technologies bloom as many as possible by providing suggestions and new perspectives for technologies that are difficult to apply.

Jianquan Liu

Since joining the company in 2012, Dr. Liu has been working on the research of this technology with Dr. Nishimura. He is primarily involved in databases and similarity search (high-dimensional) and has released behavior detection and many other technologies at NEC. Many of Dr. Liu's research papers have been published at conferences.

Q1: So how do you feel about receiving this award?
I am extremely happy. While the video search technology, which became the prototype of this awarded technology, was commercialized back in 2012, it was temporarily terminated just as we were about to ship it overseas and run for a proof of concept. At the time, customers had not accumulated any data yet, and the market had not caught up to the technology. We got too far ahead of the market and the needs. So I feel that our efforts to overcome that setback and continue striving have been rewarded.

Q2: What do you think led to this year's award?
I believe that it was driven by the concept of discovering unregistered people. Previous video search and biometric authentication focused on looking for and authenticating people who had been registered in advance. However, utilizing our technology, you can detect and connect the same people appearing multiple times in the videos of different locations even if they are unregistered. We then realized that if we utilize the appearing frequency and other information, we can observe those appearances with a certain meaning such as suspicious loitering or lost children. I believe that thinking about the technology in this way from a different angle was a significant factor to receiving the award.

Q3: Are you currently undertaking any new objectives?
I am currently pursuing research based on the belief that this technology can be applied in various fields. The technology has currently already been commercialized not only for town monitoring applications but also as a technology for searching for performers on television programs, and I believe that there are many other uses. For example, it could be applied to marketing. If a person is detected across multiple outlets of the same chain store, they may have trouble finding something. In the case of a theme park, you could recommend a frequent visitor to purchase an annual pass. I think that it can be flexibly applied according to the ideas, so I am advancing research into such applications.