Fujikura Ltd. contributes to the development of various industries, including information and telecommunications, energy, and electronics. The company is currently conducting a demonstration experiment of its "AI security system Asilla" at Fukagawa Gatharia, a commercial complex in Tokyo. We interviewed Mr. Noriyuki Karasawa, Chief Engineer of the Planning Department in the New Business Creation and R&D Division.
Planning Department, New Business Creation and R&D Division
Mr. Noriyuki Karasawa, Chief Engineer
We possess the Fukagawa Gatharia commercial complex area in Koto-ku, Tokyo and are actively undertaking multiple initiatives to elevate its worth. One of our endeavors is to bolster security by incorporating AI into our surveillance cameras. Among the diverse array of solutions available, "Asilla" captivated our attention with its distinct approach, prompting us to initiate a trial implementation approximately two years ago. Initially, we commenced with a limited number of cameras, but now we employ "Asilla" across dozens of security cameras. Although currently in the demonstration experiment phase, our ultimate aspiration is to integrate it on a comprehensive scale in due course.
The detection of abnormal behaviors by "Asilla" played a pivotal role in our decision-making process. We have not encountered similar functionality in other solutions. Conventional image recognition relies on extensive data training to teach AI to identify suspicious behaviors, but it can only detect behavior that is predictable or anticipated in advance, such as fighting or falling. Frankly speaking, we found this approach to be too limited in scope. However, "Asilla" goes beyond that. It learns normal behavior patterns and can detect even subtle deviations from them. The system's ability to identify behaviors that cannot be assumed or predefined in advance strongly influenced our decision to introduce it.
Earlier this year, I was shown many videos in which discomfort was actually detected. Even from the human point of view, the system was able to detect behaviors that were "not necessarily abnormal, but certainly abnormal. We originally decided to introduce the system because of its ability to detect discomfort, but when we saw its unique functionality, we thought, "This could be a great system. If the system can do this well at this stage, I have high hopes that it will be able to detect abnormal behavior before full-blown criminal acts in the future.
The value of real estate will decline year by year if no action is taken. The challenge is how to maintain and improve the value of Fukagawa Gatharia. Although improving the value of real estate is a difficult task, it is important to make users feel that they want to come back here again. Safety is one of the major factors in achieving this, so we put a lot of effort into security. In fact, we have deployed more security guards.
We believe there are two challenges. The first is to reduce costs, and the second is to maintain and improve security quality. We would like to focus on the latter in particular in the future. If an accident or incident were to occur, it would cause discomfort to users. Although cost reduction with easy-to-understand results tends to be praised, safety cannot be neglected. Under such circumstances, we believe that "Asilla" can be a powerful tool to prevent nuisance and criminal acts. Security guards check the images from security cameras on their monitors, but it is not possible to show all the images on the monitors at the same time. We hope to improve the quality of security by utilizing the power of AI.
While the identification of abnormal behavior serves as evidence that the AI is functioning correctly, some individuals may question whether it could be a false positive. It would be even more valuable if the system could differentiate between situations that warrant an alert and those that do not, or if it could categorize different levels of abnormality. As the ability to detect abnormal behavior is a unique feature of "Asilla," we eagerly anticipate further improvements in accuracy and enhancements to its functionality.
While it is important to detect behaviors such as falling or staggering that require a security guard to call out, the most important thing is to detect behaviors that lead to incidents and accidents, such as crimes. Of course, there may be areas that have already been detected, and we would like to see you take on the challenge of proving this. I would love to work with you on a demonstration experiment to see how far we can actually go in detecting such behavior. I am sure that experiments involving researchers in criminal psychology and human behavior will yield quite interesting results. In the future, I hope that "Asilla" will be developed into a behavior recognition AI that can prevent criminal acts before they occur, thereby contributing to the safety of Japan and the world.
Company name: Fujikura Ltd.
Location: 1-5-1 Kiba, Koto-ku, Tokyo
Representative: Naoki Okada
Founded: March 1910
Business activities: fiber optic cables, optical transmission systems, communication systems, electronic wiring components
Automotive electrical components, power systems, industrial wires, metal materials, real estate business
Official website: https://www .fujikura.co.jp/