Research On Ship Wake Detection Method In SAR Image Based On Electromagnetic Scattering Simulation

The strong background noise caused by sea clutter and the rapid development of ship stealth materials have brought great difficulties to the detection of marine ship targets in synthetic aperture radar(SAR)images.By detecting the ship wake,the relevant information of the ship can be obtained indirectly.In addition,ship wake SAR imaging based on electromagnetic scattering mechanism can not only better clarify the physical mechanism of wake imaging,but also effectively expand the sample database of ship target recognition.Therefore,the research on ship wake detection method in SAR image based on electromagnetic scattering mechanism has important research significance for target recognition and ocean remote sensing in complex marine environment.Traditional wake detection methods can not effectively detect nonlinear wake and weak wake under high sea conditions.This paper explores the methods to further solve the above problems,carries out relevant research on ship wake detection,establishes the composite scene model of the sea surface and its wake,analyzes the wake bin distribution of the sea surface and the sea background with the help of the semi deterministic bin model,and further simulates the wake SAR imaging under the sea background based on the modulation model.Combined with the simulation results and measured results of wake SAR imaging,the sample database is constructed.The YOLO(You Only Look Once)algorithm based on deep learning improves the efficiency of wake detection and effectively realizes the detection of weak wake.The main work of this paper includes:1.Aiming at the mixed waves existing in the marine environment,based on the sea spectrum model,the wind wave and swell components in the waves are described respectively,and the geometric modeling of typical wind waves and wind wave swell mixed waves is realized.Based on the mathematical model of Kelvin wake,the geometric modeling of ship wake is carried out,and on this basis,the composite scene model of sea surface and its wake is established.Based on the simulation results,the wave height and wavelength characteristics of the wake are analyzed.Combined with the semi deterministic panel model,the effects of radar parameters such as incident angle,wind speed,wind direction,polarization condition,ship speed and heading and marine environmental parameters on the backscattering coefficient and panel scattering distribution characteristics of the sea surface and its wake are analyzed.2.The principle of azimuth resolution and range resolution of synthetic aperture radar and the phase frequency characteristics of linear frequency modulation signal(LFM)are described in detail.The SAR imaging model of sea surface based on modulation mechanism is given,and the SAR image of wake under sea surface scene is simulated based on the panel scattering distribution of "sea surface + wake" composite scene.The SAR imaging simulation results of wake under different polarization,ship speed and heading are analyzed,and the parameter conditions conducive to wake recognition under the influence of different parameters are emphasized.3.Aiming at the problem that traditional wake detection methods are difficult to detect nonlinear wake and wake in strong background noise environment,the wake detection algorithm based on depth learning is used,combined with the wake SAR imaging simulation results and some measured wake SAR images,which effectively expands the sample set for depth learning,combined with the mean Average Precision(m AP)index,The model performance of the wake SAR simulation results as the test set under different sea conditions,different network structures is evaluated,and the model performance of the measured SAR wake image as the test set under different network structures is evaluated.Compared with the traditional wake detection method and the wake detection method based on depth learning,the effectiveness of the wake detection method based on depth learning for nonlinear wake and weak wake recognition is analyzed.