Borrion, H.; (2007) Study of processing techniques for radar non-cooperative target recognition. Doctoral thesis , University of London.

PDF U591312.pdf Download (26MB)

Abstract

Radar is a powerful tool for detecting and tracking airborne targets such as aircraft and missiles by day and night. Nowadays, it is seen as a genuine solution to the problem of target recognition. Recent events showed that cooperative means of identification such as the IFF transponders carried by most aircraft are not entirely reliable and can be switched off by terrorists. For this reason, it is important that target identification be obtained through measurements and reconnaissance based on non-cooperative techniques. In practice, recognition is achieved by comparing the electromagnetic sig nature of a target to a set of others previously collected and stored in a library. Such signatures generally represent the targets reflectivity as a function of space. A common representation is known as one-dimensional high-resolution range-profile (HRRP) and can be described as the projection of the reflectivity along the direction of propagation of the wave. When the measured signature matches a template, the target is identified. The main drawback of this technique is that signatures greatly vary with aspect-angle so that measurements must be made for many angles and in three dimensions. This implies a potentially large cost as large datasets must be created, stored and processed. Besides, any modification of the target structure may yield incorrect classification results. Instead, other processing techniques exist that rely on recent mathematical algorithms. These techniques can be used to extract target features directly from the radar data. Because of the direct relation with target geometry, these feature-based methods seem to be suitable candidates for reducing the need of large databases. However, their performances and their domains of validity are not known. This is especially true when it comes to real targets for at least three reasons. First, the performance of the methods varies with the signal-to-noise ratio. Second, man-made targets arc often more complex than just a set of independent theoretical point-like scatterers. Third, these targets are made up of a large number of scattering elements so that mathematical assumptions are not met. In conclusion, the physical correctness of the computational models are questionable. This thesis investigates the processing techniques that can be used for non-cooperative target recognition. It demonstrates that the scattering-centre extraction is not suitable for the model-based approach. In contrast, it shows that the technique can be used with the feature-based approach. In particular, it investigates the recognition when achieved directly in the z-domain and proposes a novel algorithm that exploits the information al ready in the database for identifying the signal features that corresponds to physical scatterers on the target. Experiments involving real targets show that the technique can enhance the classification performance and therefore could be used for non-cooperative target recognition.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as XMLCSVJSON

Archive Staff Only

View Item