The importance of nuclear security increased after FuNushima Daiichi nuclear power plant accident. Especially as a threat to nuclear power plants, sabotage by insider is worthy of attention. In response to this situation, hand motion is an important part of human activity and it has high contribution to high-accuracy detection of insiders’ sabotage. Moreover, Time series data analysis is a useful method in abnormal behavior detection. In this research, the real-time hand motion detection system was develope...
英字タイトル:
Detection of Insiders’ Sabotage using Time-Series Data Analysis of Hand Motion

Piping and Instrumentation Diagrams (P&IDs) contain information about the piping and process equipment together with the instrumentation and control devices, which is essential to the design and management of Nuclear Power Plants (NPPs). There are abundant complex objects on P&IDs, with imbalanced distribution of these objects and their linked information across different diagrams. The complexity of P&IDs thus is increased which make automatic identification difficult. Therefore, the content of P&IDs is gen...

In this research, a new method was developed to identify the _hand motion” of malicious sabotage behaviors. The Auto-encoder and Auto Associative Neural Network were applied for identification, and the time-series data of finger tips position were used as the training data of these Machine Learning. The identification reliability was more than 66%. ...
英字タイトル:
Detection of Sabotage Motion by Time Series Analysis of Video

In order to quantitatively evaluate the ability of a nuclear plant to recover its safety functions, we are developing a method to simulate accident management in chronological order according to an accident scenario, rather than simply evaluating the probability, and to evaluate whether or not a major accident will eventually occur, i.e., whether or not the minimum necessary safety functions can be recovered within a time limit. In this presentation, we will discuss the development of a method to evaluate w...
英字タイトル:
Visualizing method for resilience of nuclear power plant

Tracking human motion from video sequences is a notable technique that is used to detect anomalies in individual human behavior. Several commercially available motion capture devices are based on the use of depth cameras. However, there are a couple of problems with the use of a depth camera. Firstly, a complicated camera system is required, and secondly, the optical field of view is limited. To overcome these problems, we need a technique that can recognize human motion from wide-angle images. In this stud...
英字タイトル:
Development of Robust Keypoint Detector for Distorted Wide-Angle Images and Application to Human Motion Analysis

In this research, a new method was developed to identify the “hand behavior” of malicious sabotage behaviors. The Convolutional Neural Network (CNN) and the Long Short Term Memory (LSTM) were applied for analysis of the time-series data of hand behavior images and identification of hand behavior....
英字タイトル:
Development of Sabotage Behavior Detection by Hand Image Analysis and Machine Learning

Sabotage by malicious insider is one of significant and serious threats for nuclear security of nuclear power plants. It is difficult, however, to distinguish abnormal behaviors from normal works such as their daily maintenance activities. In this study, a technique was proposed to subdivide the abnormal behavior due to sabotage by image analysis and then to detect and identify the abnormal behavior in real time. ...
英字タイトル:
Proposal of Insider Detection Method for Nuclear Security

According to the IAEA's Incident and Trafficking Database (ITDB) [1], nuclear security-related incidents occur approximately every three days in the world, with the majority being nuclear and radioactive material theft. Generally, material accountancy is applied as a countermeasure against theft. In this study, a deep learning-based approach was proposed to identify theft of nuclear and radioactive material. As various acts can be performed at the site where nuclear material is handled, the recognition of o...
英字タイトル:
Identification of theft from normal operations in nuclear facilities

A deep learning model has been proposed to recognize hand action for nuclear security A system has been developed that can automatically recognize hand action from video data acquired by a single depth camera...
英字タイトル:
Hand Motion Recognition from Movie Data by Deep Learning

An algorithm has been developed to detect sabotage, etc. by converting surveillance camera images and rule documents into graph structures by deep learning. In the verification using the demo video, the judgment accuracy of 90% or more was obtained. A basic technology has been established to realize an interface with natural language processing AI with image AI....
英字タイトル:
Detection of malicious acts on nuclear security by combining image recognition and natural language processing