Hybrid method to extract striation features from ship noise spectrogram

摘要

The features of interference striations excited by a passing ship are strongly determined by the acoustic waveguide properties. These striation position and orientation have been used for environmental inverse problems. The ship noise spectrogram can be very noisy due to measurement conditions, i.e., high ambient noise level or transmission loss noise. It is necessary to enhance the underlying interference structure before extracting the striation features of interest. A hybrid image processing method is introduced in this paper for interference structure enhancement. It first uses a Gabor filter bank to provide the local image intensity maximum value in different directions, and then locally equalizes the resulting image. Different ship noise data sets from different experiments are processed by the proposed method. Preliminary results demonstrate that the hybrid method can effectively identify striations in both low and high frequency regions, especially for the data set collected under particularly difficult measurement conditions due to strong current, surface wave, high ambient noise level, complex time-varying source spectrum, etc. Consequently, better estimates of the position and orientation of local striations can be obtained, which will likely improve the accuracy of striation-based inversion techniques.