[en] In many Safeguards verifications it is important to detect changes made in a given installation or track the progression of the construction work in a new plant. This problem is commonly known as Design Information Verification (DIV). The process of DIV can be divided into two main steps: i) acquire raw data from the plant; ii) compare the 'as-built' and target models. One of the objectives of this work was to develop a prototype accepting multi-sensory data as input. This prototype is also multi-scale, a feature accommodating different acquisition systems and algorithms depending on the size of the objects/buildings to be modelled. Indeed, a building, large tanks or small pipes require different equipment and modelling algorithms. The system provides semi-automated tools to compare the acquired 'as-built' and the approved design models. The system is a scaled based approach combining different sensors and 3D reconstruction techniques depending on the size of the object to be modelled and accuracy of the final model. Both software and hardware are designed to cover a wide range of distances between 1mm to 200m. This includes the 3D modelling of buildings, large objects, e.g., fuel containers, large pipes and very small pipes. The paper describes the main components of the present implementation of the Design Information Verification prototype. It will concentrate on: 1. Description of the data acquisition systems 2. Tools for construct realistic 'as-built' 3D computer models of the Safeguards relevant part of the plant 3. Tools to verify the currently 3D reconstructed model with a previously acquired model or the original design model. 4. System to track and document changes in successive inspections. Currently, DIV activities are labour-intensive and are mostly carried out by visual inspection, the taking of photographs and reference to engineering drawings. In addition, they not offer the level of accuracy and conclusiveness now possible using laser range finding techniques. The technique presented in the paper effectively and conclusively record detailed constructional information in support of DIV for large and complex industrial nuclear facilities. Measurement data is gathered at an initial DIV. Those data is then 70 compared to new data taken at subsequent DIVs. Differences are thus identified and resolved in a timely manner

[en] Surveillance techniques are based on the detection of changes. These changes can be caused by moving objects or people, or by modifications made to the environment itself. Visual surveillance uses optical means, e.g., the analysis of an image acquired by a surveillance camera. These techniques are effective in detecting objects moving within the surveyed area. There are situations, however, where optical surveillance may prove to be unreliable. In some cases, the changes in the image are too small to be properly detected with scene change detectors. In other cases, alarms are generated without objects (or people) moving. These false alarms may be caused by changes in illumination, e.g., a faulty lamp or spurious reflections in places near water pools. Further, the absence of illumination during a blackout (whether it is caused by accident or on purpose) prevents cameras from their surveillance operation. There are high security installations for which it is necessary to introduce reliable, independent and effective sensors that can keep the surveillance work even during a blackout. Laser range scanners are electronic instruments measuring the distance from the instrument itself to the outside world along a specific direction. The type of the instrument to use depends on the range of distances to measure. Indeed, whereas for large distances (e.g. between 1 and 200m) it is possible to use time-of-flight instruments, for short distances (e.g., from a few centimetres to about 1.5m) a triangulation laser striping system is used. The deflection of the laser beam (e.g., using rotating mirrors) enables the acquisition of the distance profiles (or matrices) of the surrounding premises in a very short time

[en] This paper presents BRIDGE, a filtering tool to assist nuclear inspectors in the review of surveillance image streams originated from plants, installations and stores. The goal of the review is to browse surveillance streams and inspect with priority safeguards-relevant sequences ('clips'), skipping the irrelevant parts. To help on this, BRIDGE filters image streams to retrieve some relevant images per clip based on their content, and uses temporal-adjacency between images to focus the inspector's attention on the complete sequence. BRIDGE is built around: (i) its user, the inspector, who can point to BRIDGE examples of safeguards relevant and irrelevant images; (ii) image features (or descriptors of image content), such as the image color histogram; (iii) a classifier (or filter) that assigns a relevant/irrelevant label to the images in the surveillance stream on the basis of their feature description; (iv) the images' time stamps, that specify the moment in time when the images were acquired. Both the image features and the classifier parameters are selected automatically and remain completely transparent to the inspector. The only input required from him/her is the labelling of some example images in a relevance feedback setting. Therefore, it is argued that the tool is easy-to-use by people with no in-depth knowledge in image processing and pattern recognition. The use of BRIDGE is illustrated by experiments on surveillance sequences acquired in-house. Results show that BRIDGE identifies most relevant images with few false positives

[en] This paper presents a system prototype for the review of surveillance image sets aiming at supporting Safeguards inspectors in the analysis of image set acquired in nuclear installations. The tool provides a user interface to easily annotating and classifying still images as relevant and irrelevant, to automatically display small video clips showing what happened in the monitored environment immediately before and after the acquisition of the analysed image, and, based on image annotation, to automatically generate a report of the activity that took place in the monitored installation. The presented prototype is based on image retrieval techniques allowing extracting from large datasets images visually similar to a user-selected example. The novelty of the system is represented by the integration of temporal acquisition information to the image retrieval operations. The combination image retrieval and image acquisition information represents a powerful tool for an easy and accurate review of all the activities that took place in the monitored environments

[en] The identification of nuclear containers consists normally to read a set of alphanumeric characters that are engraved, embossed or flame-etched in one of the surfaces of the containers. One of the factors affecting the identification of the string of characters is the quality of the visual image. The quality is affected by differences in illumination of the scene, by spurious external illumination sources, or from the poor and inconsistent contrast between the material surface and engraved/embossed characters. Laser-scanning technologies alone or combined with a camera provide a robust and efficient way for reading and automated recognition of identification numbers. Further, this technique does not have the problems of camera based reading techniques, i.e., complex illuminations systems and difficult set-up procedures. The paper describes three working prototypes of an automated system reading the identification string of nuclear containers (waste drums or fuel elements). A single, integrated system accepts a container, performs a 3D and visual scanning, identifies its type (embossed, engraved or flame-etched) and uses the combination of the 3D and visual data to automatically read the identification string. The processing cycle and the software tools and algorithms developed for the automatic recognition of identification strings are also described. The system automatically detects the 3D position of the strings and segments the different characters either based on the 3D information (engraved or embossed characters) or by combining an image from a camera with the range information (flame-etched characters)

[en] Virtual Private Networks (VPN) is an important technology allowing for secure communications through insecure transmission media (i.e., Internet) by adding authentication and encryption to the existing protocols. This paper describes some VPN performance indicators measured over international communication links. An ISDN based VPN link was established between the Joint Research Centre, Ispra site, Italy, and EURATOM Safeguards in Luxembourg. This link connected two EURATOM Safeguards FAST surveillance stations, and used different vendor solutions hardware (Cisco router 1720 and Nokia CC-500 Gateway). To authenticate and secure this international link, we have used several methods at the different levels of the seven-layered ISO network protocol stack (i.e., Callback feature, CHAP - Challenge Handshake Protocol - authentication protocol). The tests made involved the use of different encryption algorithms and the way session secret keys are periodically renewed, considering these elements influence significantly the transmission throughput. Future tests will include the use of a wide variety of wireless media transmission and terminal equipment technologies, in particular PDAs (Personal Digital Assistants) and Notebook PCs. These tests aim at characterising the functionality of VPNs whenever field inspectors wish to contact headquarters to access information from a central archive database or transmit local measurements or documents. These technologies cover wireless transmission needs at different geographical scales: roombased level Bluetooth, floor or building level Wi-Fi and region or country level GPRS

[en] There are situations, where video surveillance based on automated scene change detection systems may prove to be ineffective. In some cases, the changes in the image are too small to be properly detected. In others, false alarms are generated without objects (or people) moving. These alarms may be caused by changes in illumination, e.g., a faulty lamp or spurious reflections in places near water pools. Further, the absence of illumination during a blackout (whether it is caused by accident or on purpose) prevents cameras from their surveillance operation. There are high security installations for which it is necessary to introduce reliable, independent and effective sensors that can keep the surveillance work even during a blackout. Laser range scanners are electronic instruments measuring the distance from the instrument itself to the outside world along a specific direction. The type of the instrument to use depends on the range of distances to measure. The deflection of the laser beam (e.g., using a rotating mirror) enables the acquisition of the distance profiles of the surrounding premises in a very short time. Alarms are generated by comparing acquired distance measurements (organized as profiles - 2D Laser Surveillance, or in clouds of points - 3D Laser Surveillance) with previously acquired references. This is most effective for real-time detection of even small changes in the environment. The main feature of laser surveillance system is that it measures distances between the instrument itself and the real world, rather than measuring only appearance as video surveillance systems do. As such, any physical change occurring within the field of site of the laser instrument is immediately measured and detected. The fact that laser systems are self-illuminating and do not depend on the external illumination make them quite effective to work during blackouts. The paper introduces laser surveillance and shows how such systems complement well already existing video surveillance systems. (author)

[en] To ensure the peaceful intent for production and processing of nuclear fuel, verifiable process monitoring of the fuel production cycle is required. As part of a U.S. Department of Energy (DOE)-EURATOM collaboration in the field of international nuclear safeguards, the DOE Sandia National Laboratories (SNL), the European Commission Joint Research Centre (JRC) and Directorate General-Energy (DG-ENER) developed and demonstrated a new concept in process monitoring, enabling the use of operator process information by branching a second, authenticated data stream to the Safeguards inspectorate. This information would be complementary to independent safeguards data, improving the understanding of the plant's operation. The concept is called the Enhanced Data Authentication System (EDAS). EDAS transparently captures, authenticates, and encrypts communication data that is transmitted between operator control computers and connected analytical equipment utilized in nuclear processes controls. The intent is to capture information as close to the sensor point as possible to assure the highest possible confidence in the branched data. Data must be collected transparently by the EDAS: Operator processes should not be altered or disrupted by the insertion of the EDAS as a monitoring system for safeguards. EDAS employs public key authentication providing 'jointly verifiable' data and private key encryption for confidentiality. Timestamps and data source are also added to the collected data for analysis. The core of the system hardware is in a security enclosure with both active and passive tamper indication. Further, the system has the ability to monitor seals or other security devices in close proximity. This paper will discuss the EDAS concept, recent technical developments, intended application philosophy and the planned future progression of this system.