Wavelet-based progressive classification of high-range-resolution radar returns

Abstract

We investigate the problem of fast and accurate classification of high range resolution radar returns from ships. In addition we investigate the problem of efficient organization of large databases of pulsed high resolution radar returns from multiple targets in order to economize memory requirements and minimize search time. We use synthetic radar returns from ships as the experimental data. We develop a novel algorithm for hierarchically organizing the database, which utilizes a multiresolution wavelet representation working in synergy with a Tree Structured Vector Quantizer (TSVQ), utilized in its clustering mode. The tree structure is induced by the multiresolution decomposition of the radar returns. The TSVQ design algorithm is ofthe "greedy" type. We demonstrate that our algorithm automatically computes the aspect graph (i.e. the simultaneous representation of compressed pulses as functions of aspect and elevation) for a single target or for a group of targets. We also develop a novel optimization framework for the simultaneous design of the wavelet basis, the Tree-Structured Vector Quantizer and the Classification rule. We describe an efficient and promising implementation consisting of an adaptive Wavelet Transform -Tree Structured Vector Quantization with Learning. We show experimental results which indicate that the combined algorithm executes orders of magnitude faster data search time, with negligible performance degradation (as measured by rate-distortion curves).

<title>Wavelet-based hierarchical organization of large image databases: ISAR and face recognition</title>

“…In this paper we use as target models scale space aspect graphs constructed using the two basic techniques mentioned above (i.e. wavelets and hierarchical clustering) as described in our earlier work [1,3,4,5]. The fundamental extension here is image data.…”

“…In Figure 1 below we show the intuitive explanation of a multi-resolution aspect graph, from our previous work on pulsed radar data [1,3,4,5]. We extend this construction here to ISAR data.…”

“…Our algorithm combines a wavelet preprocessor (in two dimensions) followed by a Tree-Structure-Vector Quantization (TSVQ) exactly as developed in [1,3,4,5]. The only difference here is that the signals (or inputs to the algorithm) are images.…”

“…The process is repeated until we reach the desired overall fidelity in the representation. This process and the associated TSVQ algorithm is described in our earlier work [1,3,4,5] and is illustrated in Figure 2. We call the resulting algorithm the Wavelet --TSVQ (WTSVQ) algorithm.…”

“…We have previously developed and applied such methods to one-dimensional signals (radar pulses, or acoustic returns) with success [1,3,4,5]. In the present paper we extend these methods to images.…”