Research On Line Spectrum Detection Technology Of Underwater Acoustic Target Radiated Noise Signal

Implementation of acoustic target detection under the complex background of the marine environment is an important issue in the field of Marine research.However,with the increase of types of underwater acoustic targets and the decrease of signal-tonoise ratio,the difficulty of target detection is higher and higher.Line-spectrum composition of ship-radiated noise with high energy intensity,small transmission loss,stable phase and obvious physical characteristic,very suitable for low signal-to-noise ratio and remote passive detection,and can be used as the main basis of target detection.This paper conducts research on line-spectrum detection,exploring line spectrum detection from both frequency domain and time-frequency domain perspectives.First of all,the generation mechanism of ship-radiated noise was studied,and a simulation method for ship radiated noise was proposed.First according to the actual ship-radiated noise power spectrum of the structure design of FIR filter,and the filter is used to filter Gaussian white noise to achieve continuous-spectrum simulation.Then,line-spectrum simulation is achieved through line-spectrum extraction and reconstruction.Finally,the simulation signal of ship-radiated noise is obtained by adding continuous-spectrum and line-spectrum.This method retains the structural characteristics of the measured signal while also having the advantage of adjustable parameters,which can meet the needs of simulation signals in different situations.The implementation principle and steps of frequency domain line-spectrum detection were elaborated.Frequency domain line-spectrum detection is based on power spectrum,and line-spectrum is determined through threshold and line-spectrum shape characteristics.Introduced commonly used classical and modern power spectrum estimation methods,and compared the algorithms under different signal parameters.Finally,considering the stability and accuracy requirements,it is believed that the Welch power spectrum estimation method is the most suitable method for autonomous line-spectrum detection.In order to adapt to the variability of environmental noise in autonomous detection,this paper also introduces four continuous-spectrum smoothing techniques,which can adaptively perform continuous spectrum fitting based on the power spectrum estimation results to reduce the impact of background noise on linespectrum detection.In response to the problem of poor performance in line-spectrum discrimination using single moment energy values in frequency domain line-spectrum detection at low signal-to-noise ratios,this paper introduces a line-spectrum detection method based on time-frequency joint detection,which transforms one-dimensional frequency point detection into line-spectrum trajectory detection on two-dimensional LOFAR images.Two line-spectrum trajectory detection methods are implemented through Canny edge detection and HMM model in image processing.The research results show that the edge detection algorithm is simple,fast,and has good detection performance when the signal-to-noise ratio is high,but the detection effect is poor for low signal-to-noise ratio and fracture line-spectrum.The HMM based line-spectrum tracking algorithm can achieve line-spectrum trajectory detection for straight lines,oblique lines,curve lines,and multiple line spectra,and also has good detection performance at low signal-tonoise ratios.A line-spectrum target detection software was developed using Qt design,achieving the integration and visualization of detection algorithms.The software was tested using two methods: self-generated simulation signals from the software and reading sea trial experimental data.The results showed that the overall operation of the software was good,and line-spectrum detection and extraction could be achieved within allowable errors,verifying the feasibility and reliability of the software.