Technology Trend “How AI is changing safety”

1. Safety inspection and AI

Due to the recent social situation, in places where many people and things gather, such as public institutions, shopping malls, and large events, ensuring safety is a duty not only for the police but also for the organization that operates it.

At airports where people and goods of various backgrounds come and go, various safety inspections are conducted daily. Since the number of passengers are expected to continuously increase, effective measures are required to maintain the quality of safety inspections while coping up with the increasing number of passengers and cargos.

AI is now being used to assist and automate some work in safety inspections. Face recognition technology using AI has already been put into practical use at several airports in the world, and the inspection load has been reduced significantly. For X-ray inspection of the cargo, technological development using AI has been promoted worldwide mainly by national institutions, it is about to be put into practice now.^*1

2. X-ray luggage inspection by AI

The X-ray inspection at airports is aimed to detect things that can be dangerous to the operation of aircraft such as explosive devices or sharp objects such as knives. In some countries, inspections are performed at the time of entry as well in order to detect any illegal goods brought into the country. NTT DATA has been developing and verifying AI technology in this area. X-ray images are usually very different from what we see through our eyes, and since the objects are projected in a special colour and transparent, it is not an easy task for human beings to judge the contents. When it comes to AI-based image analysis, you might think of deep learning, and deep learning is highly effective in special images such as X-ray. Deep learning has a technique called “object detection” which allows you to estimate the type and position of a specific object from the image. It is generally used as a method for detecting cars and people from natural images, but by making it learn characteristics of X-ray images, NTT DATA in-house verification confirmed that X-ray images can be detected with high accuracy.

3. Challenges for practical use

However, when the AI technology was verified and when the accuracy was checked, the same was applied to the luggage inspection of actual passengers, the outcome was not as expected. The actual passengers come from various countries with different backgrounds, compared to experimentally created data, a wide variety of the goods packed in various ways were far beyond recognition. As a result, it caused discrepancies and many things looked overlapped then overlooked. When AI is applied under real world scenarios, there are many situations that were not anticipated in studies. This is not limited to X-ray analysis but for the practical application of AI, it is required to repeat the verification process in the actual fields and reflect the feedback of the results in order to make it into practical use. NTT DATA will continue to create opportunities for verification on a global scale in order to ensure the technology is put into practical use and try and contribute to the safety of the world.

• *1Kevin J. Liang; Geert Heilmann; Christopher Gregory; Souleymane O. Diallo; David Carlson; Gregory P. Spell; John B. Sigman; Kris Roe; Lawrence Carin, “Automatic threat recognition of prohibited items at aviation checkpoint with x-ray imaging: a deep learning approach”