Denis Kolev

Symbol recognition with a new autonomously evolving classifier autoclass

A new algorithm for symbol recognition is proposed in this paper. It is based on the AutoClass classifier [1], [2], which itself is a version of the evolving fuzzy rule-based classifier eClass [3] in which AnYa[1] type of fuzzy rules and data density are used. In this classifier, symbol recognition task is divided into two stages: feature extraction, and recognition based on feature vector. This approach gives flexibility, allowing us to use various feature sets for one classifier. The feature extraction is performed by means of gist image descriptors[4] augmented by several additional features. In this method, we map the symbol images into the feature space, and then we apply AutoClass classifier in order to recognise them. Unlike many of the state-of-the-art algorithms, the proposed algorithm is evolving, i.e. it has a capability of incremental learning as well as ability to change its structure during the training phase. The classifier update is performed sample by sample, and we should not memorize the training set to provide recognition or further update. It gives a possibility to adapt the classifier to the broadening and changing data sets, which is especially useful for large scale systems improvement during exploitation. More, the classifier is computationally cheap, and it has shown stable recognition time during the increase of training data set size that is extremely important for online applications.

ARFA: Automated real-time flight data analysis using evolving clustering, classifiers and recursive density estimation

In this paper a novel approach to autonomous real time flight data analysis (FDA) is proposed and investigated. The anomaly detection is based on recursive density estimation (RDE) and the fault identification is based on the evolving self-learning classifiers introduced recently. The paper starts with a brief critical analysis of the currently used FDA methods and tools. Then the problems of fault detection (FD) and identification are described formally. The importance of the ability to process the data in real time and on-line (in flight) is directly related to the efficiency and safety. Therefore, in this paper the focus is on the recursive approaches which are computationally lean and suitable for on-line mode of operation. The novel concept of ARFA (Automated Real-time FDA) is then applied to real flight data from Russian and USA made aircrafts. The results are compared and analyzed. Both, advantages that this novel methodology and algorithms offer as well as the current limitations and future directions of research are pointed out and future work outlined.

Security situation management - developing a concept of operations and threat prediction capability

This paper addresses a collaborative security situation management capability for air navigation. In particular, we formulate the development of a threat prediction capability as a situation management problem mapping the concepts of situation awareness and information fusion. Air transportation and air navigation is undergoing a fundamental transformation. This also requires novel approaches to system security and the management of security incidents across a network of actors. The Global ATM Security Management project addresses this problem space. The work reported in this paper, conceptualizes a security function that supports the management of security incidents on a local, national, and regional level supporting the collaborative effort of classical air traffic management stakeholders and security stakeholders. The security function is based on a network of distributed nodes and capabilities. One such a capability is the threat prediction model. This component is based on a representation of the (sub-) system context as a network of supporting assets, event detection sensors, and associated security controls. Based on the description of the (sub-)system context as a sequence of situations, the threat prediction capability addresses the identification of a security incident and its potential impact as an optimization problem. This paper reflects the work of the first year of the project. In particular, it demonstrates the general feasibility of the approach and the further modelling and preparatory work for further validation activities.

Towards validating a security situation management capability

With SESAR and NextGen readying towards implementing novel operational concepts and technical enablers in ATM/CNS, the question of how to manage security in a dynamic environment across a highly distributed and networked system gains higher attention. The Global ATM Security Management project (GAMMA) addresses the development of such a security situation management capability. Following the September 11 attacks and major large-scale outages of critical infrastructures, the security of air navigation has emerged as a critical capability gap. On-going transformation programs like SESAR and NextGen are moving into the deployment phase with limited to none tangible security solutions. GAMMA addresses this gap by investigating a security situation management capability. The framework of this capability is devised as a distributed network of aviation stakeholders that jointly collaborate in identifying and localizing security incidents while considering the constraints given by the different participants, national responsibilities, and collaboration-related requirements. This paper addresses the preparatory work for the validation of an initial security situation management capability. For that purpose, project partners setup a joint configuration and trial network for the security functions and systems developed in the frame of a real-time human-in-the-loop simulation. The simulation results have been measured against the mapping of the operational concept and validation requirements, in particular in terms of situational awareness on the operator side and networked incident management response. These results will inform the further validation activities of the project.

Incremental Anomaly Identification in Flight Data Analysis by Adapted One-Class SVM Method

In our work we used the capability of one-class support vector machine (SVM) method to develop a novel one-class classification approach. Algorithm is designed and tested, aimed for fault detection in complex technological systems, such as aircraft. The main objective of this project was to create an algorithm responsible for collecting and analyzing the data since the launch of an aircraft engine. Data can be transferred from a variety of sensors that are responsible for the speed, oil temperature and etc. In order to provide high generalization level and sufficient learning data sets an incremental algorithm is considered. The proposed method analyzes both “positive”/“normal” and “negative”/ “abnormal” examples. However, overall model structure is based on one-class classification paradigm. Modified SVM-base outlier detection method is verified in comparison with several classifiers, including the traditional one-class SVM. This algorithm has been tested on real flight data from the Western European and Russia. The test results are presented in the final part of the article.

Multiple video object tracking using variational inference

In this article a Bayesian filter approximation is proposed for simultaneous multiple target detection and tracking and then applied for object detection on video from moving camera. The inference uses the evidence lower bound optimisation for Gaussian mixtures. The proposed filter is capable of real time data processing and may be used as a basis for data fusion. The method we propose was tested on the video with dynamic background,where the velocity with respect to the background is used to discriminate the objects. The framework does not depend on the feature space, that means that different feature spaces can be unrestrictedly used while preserving the structure of the filter.

Mathematical Modelling in Air Traffic Management Security

This paper addresses the potential of mathematical modelling in support of the classical security risk assessment and treatment approach. Classical security risk assessment and control selection is strongly based on expert judgment. Within the context of large scale system implementation in air traffic management, there is only a limited availability of resources during the system engineering phase. From that perspective an alternative approach based on system engineering artefacts is highly desirable. Furthermore, robust mathematical modelling can support in the verification of security risk mitigation decisions and provide a means to address trade-off decisions between a variety of different security controls. The research reported in this paper is based on game-theoretic concepts and graph theory. The security control selection problem is modelled as a multi-objective optimization problem. Two interwoven models are developed for addressing the security risk assessment problem of a system. The internal model describes the actual system and its parameters, while the external model is used to describe possible threat scenarios. These models and the modelling technique is instantiated for a simple airport context, and the essential building blocks of the method are discussed on this example. The work reported in this paper shows the general feasibility of a mathematically founded approach to security risk assessment in large-scale system engineering. The proposed modelling approach forms the basis for the development of a dynamic security risk management capability as part of a recently started European research project on global air traffic management security.

Incremental anomaly identification by adapted SVM method

In our work we used the capability of one-class support vector machine (SVM) method to develop a novel one-class classification approach. Algorithm is designed and tested within the project SVETLANA aimed for fault detection in complex technological systems, such as aircraft. The main objective of this project was to create an algorithm responsible for collecting and analyzing the data since the launch of an aircraft engine. Data can be transferred from a variety of sensors that are responsible for the speed, oxygen level etc. In order to apply real time (in flight) application a recursive learning algorithm is proposed. The proposed method analyzes both “positive”/”normal” and “negative”/ “abnormal” examples The overall model structure is the same as an outlier-detection approach. The most important benefits of the new algorithm based on our algorithm are verified in comparison with several classifiers, including the traditional one-class SVM. This algorithm has been tested on real flight data from the USA, Western European as well as Russia. The test results are presented in the final part of the article.

A novel approach to estimating current and future states of technical systems

Modern technical systems become more sophisticated and consist of many components. It is getting harder and harder to keep the system in operational status. One of the most important tasks is to evaluate the current state of the system. The most commonly used method is the monitoring of all system parameters. This is a challenging task as each system component provides numerous parameters and all of them should be taken into account for providing accurate estimation of current and future state of the system. In the first part of the paper we introduce and evaluate a novel concept of overall state of the system called Complex Technical System Health (CTSH). In the second part of the paper we discuss implementation issue, associated with the hardware configuration of a system components in the context of distributed storage.