Micro-Doppler Based Recognition of Ballistic Targets Using 2D Gabor Filters

Abstract

The capability to recognize ballistic threats, is a critical topic due to the increasing effectiveness of countermeasures and to economical constraints. In particular the ability to distinguish between warheads and decoys is crucial in order to mitigate the number of shots per hit and to maximize the ammunition capabilities. For this reason a reliable technique to classify warheads and decoys is required. In this paper the use of micro-Doppler signatures in conjunction with the 2-Dimensional Gabor transform is presented for this problem. The effectiveness of the proposed approach is demonstrated through the use of real data.

On Model, Algorithms, and Experiment for Micro-Doppler-Based Recognition of Ballistic Targets

“…Then, the resulting matrix¯ (ν, ε) is filtered with a bank of Gabor filters whose impulse responses are The selection of these parameters has to be conducted in order to get an accurate representation of the image under test. In fact, since by varying θ m , the harmonic response of the filter moves on a circumference, whose radius is f l , it is possible to extract local characteristics in the Fourier domain by choosing a set of values for the two parameters [21]. The value of each pixel of the output image is given by the convolution product of the Gabor function and the input image¯ (ν, ε) as…”

“…The first approach is based on the statistical characteristics of the unit area function obtained by averaging and normalizing the CVD (ACVD). The second method is based on the use of pseudo-Zernike (pZ) moments [5], [18]- [20], and the third one is based on the use of the Gabor filter [21]. The ACVD approach is known to require less computation compared to the other two methods, since a smaller feature vector dimension is used.…”

“…The last two types of features are selected for their high accuracy of performance. Moreover, since a work on successfully employment of Gabor filter for ballistic target classification has been presented in [21], the methods is taken as term of comparison extending the previous work with the application on simulated data and Booster data.…”

“…Finally, the outputs of the filters are processed to extract the feature vector used to classify the targets. In particular, a feature is extracted from the output image of each filter by adding up the values of all pixels[21], as…”

Micro-Doppler classification of ballistic threats using Krawtchouk moments

“…The micro-Doppler signature has been used in [8] to extract feature based on the pseudo-Zernike moments for target classification. This approach has been also used successfully to recognize ballistic threats in [9], in which other two different feature extraction approaches are analysed: the 2D (2-dimensional) Gabor filters based features [10] and the Average Cadence Velocity Diagram (ACVD) based features. These three techniques are based on the processing of the Cadence Velocity Diagram (CVD) which represents the cadence of each signal frequency component.…”