基于分数谱时频特征的SAR目标检测与识别方法研究

发布时间：2018-04-15 10:06

本文选题：SAR图像 + 最优分数域时频谱分析　；参考：《电子科技大学》2014年硕士论文

【摘要】：SAR目标自动识别(Automatic Target Recognition,ATR)是SAR图像解译和分析的重要环节。由于SAR图像不同于其它成像图像,它掺杂着大量的相干斑噪声、几何畸变,易受方位角影响,从而导致背景散射与目标散射间存在差异。因此,如何从SAR图像中检测出目标区域,并提取目标有效信息,从而对SAR目标进行合理解译是一个亟待攻克的难题。本文主要研究了SAR图像最优分数域时频特征的提取,以及在此基础上完成的高精度SAR目标检测与识别方法。本文在相关科研工作者研究的基础上,将时频分析与分数阶傅里叶变换相结合,增加了这种瞬态分析方法的灵活性,并将该方法用于SAR地面目标检测,进而开展目标识别方法研究。本文所做的主要研究工作和贡献如下:(1)研究了SAR ATR的基本理论,掌握了SAR图像的特性,包括其成像机理,噪声产生原理及抑制,对SAR目标特性也进行了分析,为本文SAR ATR系统的构建打下理论基础。(2)学习了时频分析、分数阶傅里叶变换(Fractional Fourier Transform,Fr FT)的基本理论,并在此基础上深入研究了分数域时频分析方法。Fr FT结合计算复杂度低的Gabor变换提出了分数阶Gabor变换(Fractional Gabor Transform,Fr GT),并把它推广至2维SAR图像信号处理中。(3)对Fr GT最优化实现进行了深入的研究,主要是针对窗函数设计和最优变换阶次2方面,并进行了实验仿真与分析。(4)为了把时频谱降维可视化,深入研究了最优实现后的时频谱的特征提取方法,选择最优特征,保证检测的精度。从而提出了基于最优分数域时频谱特征的SAR目标检测方法,自适应最优分数阶Gabor变换(Optimal Fractional Gabor Transform,OFr GT),并进行了实验仿真与对比分析。(5)深入研究了分类模块的设计方法,对分类策略及特征组合进行了讨论,参数寻优以及决策融合方式也进行了深入讨论,进行了实验仿真与分析。对实际MSTAR图像库处理的结果中表明,分数域时频分析方法为SAR图像处理提出了一个新思路,且本文提出的检测和识别方法是可行的。
[Abstract]:Automatic Target recognition (ATR) is an important part of SAR image interpretation and analysis.Because the SAR image is different from other imaging images, it is mixed with a lot of speckle noise, geometric distortion and easy to be affected by the azimuth, which leads to the difference between the background scattering and the target scattering.Therefore, how to detect the target area from the SAR image and extract the effective information of the target so as to interpret the SAR target reasonably is a difficult problem to be solved.This paper mainly studies the extraction of time-frequency features in the optimal fractional domain of SAR images and the method of high precision SAR target detection and recognition based on this.In this paper, the time-frequency analysis and fractional Fourier transform are combined to increase the flexibility of this transient analysis method, and the method is applied to SAR ground target detection.Then the research of target recognition method is carried out.The main research work and contributions of this paper are as follows: 1) the basic theory of SAR ATR is studied, and the characteristics of SAR image are mastered, including its imaging mechanism, noise generation principle and suppression. The characteristics of SAR target are also analyzed.The basic theory of time-frequency analysis and fractional Fourier transform (Fractional Fourier transform) is studied for the construction of SAR ATR system in this paper.On this basis, the fractional domain time-frequency analysis method. Fr FT combined with the Gabor transform with low computational complexity is proposed. The fractional Gabor transform Fractional Gabor Transformator Fr GTN is extended to 2D SAR image signal processing.The realization is studied deeply.Aiming at the window function design and the order of optimal transformation, the experiment simulation and analysis are carried out. In order to visualize the dimensionality reduction of the time-frequency spectrum, the feature extraction method of the time-frequency spectrum after the optimal realization is deeply studied, and the optimal feature is selected.Ensure the accuracy of the detection.Thus, a SAR target detection method based on the spectrum features in the optimal fractional domain is proposed, and the adaptive optimal fractional Gabor transform is optimized Fractional Gabor transform OFR GTN. The design method of the classification module is deeply studied by the experimental simulation and comparative analysis.Classification strategy and feature combination are discussed, parameter optimization and decision fusion are also discussed, and experimental simulation and analysis are carried out.The results of actual MSTAR image database processing show that the fractional domain time-frequency analysis method provides a new idea for SAR image processing, and the detection and recognition method proposed in this paper is feasible.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【相似文献】