Neural approaches to ship target recognition

Scale and shift invariant time/frequency representation using auditory statistics: Application to rhythm description

“…It is extremely useful for, though not limited to, image processing. It has been introduced by the optical research community in the end of the seventies [3,26] and has been used in many fields since then: radar 2D signal analysis [11,14], pattern recognition in image [23,10]. It has never been used however for audio signal description.…”

Civilian vessel classification with COSMO-SkyMed images based on feature analysis

“…The formula for the kernel density estimator is shown as the (2) j (x, h) � (nh) ' t, K r x � X,) } (2) where {Xn} are the sample pixels of the vessel. K is a function satisfying J K(x)dx = 1, which is called the kernel, and h is a positive number, usually called the bandwidth or window width.…”

“…Inggs et al used Neural Approaches to classify vessels based on the feature vectors extracted by Mellin Transform [2]. Musman et al analyzed the location, size and shape of the superstructure to obtain classification results [3].…”

The Mellin Matched Filter

“…In combination with the Fourier transform invariance to signal delays, the Fourier-Mellin transform, proposed in [8], yields a representation of a signal that is independent of delay and scale change. Implementation of the Fourier-Mellin transform was suggested and studied in image processing and pattern recognition [22]- [27]. Later the Mellin transform was used as a very helpful instrument to solve different problems of signal processing: wavelet filtering [28], scale representation [7], accuracy analysis of the velocity estimation [17], computation of affine time-frequency distributions [18], etc.…”