Broad-band sound field excited by a moving source usually exhibit striations, whose structures are characterized by the propagation medium. Such relationship has been explored for underwater inverse problems, including sediment geoacoustic characterization, source localization and target recognition. The primary step for these applications is feature extraction by interference structure processing. Image processing methods based on the Radon transform, Hough transform and two-dimensional fast Fourier transform (FFT) have been used to process the interference structure to estimate the overall slope of the striations. These transforms do not allow extracting local properties of the striation pattern, e.g., the striation positions, which are also closely related to environment properties. In this paper, a multi-scale line filter based on the image Hessian matrix eigenvalues analysis is introduced and applied in searching for geometrical structures that can be regarded as a line. The location and slope of specific striation can be easily estimated from the filtered image for further applications. The possibility of using a set of striation positions for the geoacoustic characterization of a layered ocean bottom is discussed in this paper.
展开▼
