Subspace-based methodologies for the non-cooperative identification of aircraft by means of a synthetic database of radar signatures

Abstract

One of the current concerns in the actual world of aviation is the lack of quick and reliable identification of any observed object unambiguously, at long distances and under any weather conditions. Thanks to the advances in modern radar technologies, these requirements have been achieved. Indeed, radars are the most suitable sensors for a rapid and reliable recognition of targets as they can operate in scenarios where visibility is very poor, such as bad weather conditions, smoky and dusty environments, etc. Target recognition using radar sensors can be divided into two techniques: cooperative and noncooperative. Cooperative techniques, known as identification friend or foe (IFF), require the communication between target and radar, while non-cooperative techniques, so-called non-cooperative target identification (NCTI), do not establish any communication with them but rely on the comparison of the measured targets with a reference database. IFF systems are Question-Answer schemes, where the unidentified aircraft receives a question from the interrogator system in the form of an encrypted signal and the target will be able to decipher it in case of being a friend, otherwise it will be categorized as foe. These systems are operative nowadays, however, the need of establishing direct communication with the target is a weak point of this technology. Along the years, cases where IFF systems did not work properly have caused tragedies that have highlighted the need for alternatives to these cooperative systems. Non-cooperative identification systems provide, in principle, enough information to identify targets without their cooperation, even if they are not aware of being observed. These non-cooperative systems take advantage of high resolution radar data. To high resolution radars, a target is composed of several points that reflect the emitted radar signal, therefore, high resolution radars provide an image of the reflectivity of a target, that is, its signature. By comparison of the acquired signature with a database of known potential targets, the identification systems apply a number of algorithms in order to compare the observed target data with this database and to determine its class. One of the most challenging aspects is the question of the assembly and update of the database of potential targets. Ideally, the database should be populated with measurements of real target signatures, nevertheless in order to guarantee a reliable recognition, it implies the previous collection of a vast number of signatures of the majority of the existing aircraft. Unfortunately, the main drawback of this approach is the unavailability of these radar signatures for neutral and adversary assets. Thus, this thesis proposes the use of a database of ideal targets generated by electromagnetic simulations. Electromagnetic prediction codes provide a relatively cheap way of generating target signatures of any target in any aspect angle and configuration so the database can be as wide as required, the database population would be fast and low-cost and its update would just imply the modelling of new targets and simulation instead of planning expensive and lengthy cooperative measurement campaigns. Additionally, since finding datasets of actual signatures is not straightforward, this approach is also very useful for the training and assessment of new algorithms. This thesis has as a main objective the development of efficient algorithms for aircraft identification in a non-cooperative way, with high success rates and employing a database populated with information from electromagnetic simulations. The proposed scenario consists in the comparison of actual signatures with a database of simulated ones, this way, a more real scenario is emulated where the collected signatures may not have the same high SNR as the ones in the database of potential targets.