基于SAR图像的舰船目标检测算法的研究

发布时间：2018-01-04 20:04

本文关键词：基于SAR图像的舰船目标检测算法的研究　出处：《大连海事大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来,合成孔径雷达成像技术与应用得到快速发展,在导航雷达存在探测盲区而且AIS不能提供非合作舰船信息的情况下,合成孔径雷达提供了另一舰船探测的重要途径,但复杂背景下的SAR图像目标检测问题一直未能得到很好的解决。本文以复杂背景下的SAR图像中舰船目标的检测算法为研究重点,给出一种解决方法,引入视觉注意机制,再结合Shearlet变换检测方法,实验结果证明取得了较好的检测效果。所做主要工作如下:对复杂背景下的海杂波的统计分布特性展开研究。介绍了几种海杂波统计分布模型及其参数估计方法,对背景杂波进行分布拟合实验,并进行分析总结。将视觉显著注意模型与CA-CFAR检测相结合,用于舰船目标的预筛选。选取SR视觉注意模型计算显著区域,生成显著图,在显著图上进行CFAR检测,得到初步筛选结果。研究了 SR模型,分析SAR图像的平均对数幅度谱曲线,阐明了模型在SAR图像上应用的可行性。结合原图与显著图的灰度直方图,从原理上说明了经过显著区域计算之后,原图的杂波分布发生变化,海杂波强度集中分布到低强度值区域,引入评价杂波强度和不均匀程度即复杂度的指标均值和标准差。结果表明,显著图的杂波复杂度指标均值和标准差均大大降低。针对一次显著图未排除的杂波干扰,本文提出求取二次显著图,在二次显著图上进行CFAR检测,实验结果显示排除了杂波干扰。针对因显著图带来的虚警现象,采用Shearlet变换检测算法来进行虚警判别。Shearlet变换检测方法在强杂波边缘容易出现虚警而且对图像大小有限制,而本文用其来进行虚警鉴别,使检测区域限制在CFAR检测结果的局部小区域内,并针对复杂背景情况给出改进方法。改进方法对系数依据阈值进行相应削弱或增强处理之后,在各方向上进行尺度间的相关操作,再进行方向上系数融合。通过对比度参数评价改进方法与前人方法,结果显示改进方法得到较大的对比度值。
[Abstract]:In recent years, synthetic Aperture Radar (SAR) imaging technology and applications have been rapidly developed, in the case of navigation radar detection blind area and AIS can not provide non-cooperative ship information. Synthetic Aperture Radar (SAR) provides another important approach to ship detection. However, the problem of target detection in SAR images with complex background has not been solved well. In this paper, we focus on the detection algorithm of ship targets in SAR images with complex background, and give a solution. Visual attention mechanism is introduced and combined with Shearlet transform detection method. The experimental results show that good detection results have been achieved. The main work is as follows:. The statistical distribution characteristics of sea clutter in complex background are studied. Several statistical distribution models of sea clutter and their parameter estimation methods are introduced. The distribution fitting experiment of background clutter was carried out and analyzed and summarized. Visual attention model was combined with CA-CFAR detection. The SR visual attention model is selected to calculate the significant area, and the salience map is generated. The preliminary screening results are obtained by CFAR detection on the salient map. The SR model is studied. The average logarithmic amplitude spectrum curve of SAR image is analyzed, and the feasibility of applying the model to SAR image is illustrated. Combining with the gray histogram of original image and salient image, the calculation of significant region is explained in principle. The clutter distribution of the original map is changed and the sea clutter intensity is concentrated in the low intensity region. The index mean and standard deviation of the complexity are introduced to evaluate the clutter intensity and the degree of inhomogeneity. The clutter complexity index mean and standard deviation of salient map are greatly reduced. In view of the clutter interference which is not excluded from the first significant map, this paper proposes a quadratic significant map to detect CFAR on the quadratic salience map. The experimental results show that clutter interference is eliminated. Using Shearlet transform detection algorithm to detect false alarm Shearlet transform is easy to appear false alarm at the edge of strong clutter and is limited to image size. In this paper, we use it to identify false alarm, so that the detection area is limited to the local small area of CFAR detection results. An improved method is given for complex background. The improved method can weaken or enhance the coefficients according to the threshold, and then operate the correlation between scales in each direction. Then the directional coefficient fusion is carried out. The improved method of contrast parameter evaluation is compared with the previous methods. The results show that the improved method has a large contrast value.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：E925;TN957.52

【相似文献】