基于局部重要性采样的SAR图像纹理特征提取方法

发布时间：2018-02-09 14:43

本文关键词： 纹理特征合成孔径雷达局部二进编码重要性采样　出处：《自动化学报》2014年02期 　论文类型：期刊论文

【摘要】：合成孔径雷达(Synthetic aperture radar,SAR)图像因为相干斑现象和目标响应的空间变化呈现出一种纹理特性,局部二进编码等局部图像特征在光学纹理描述中获得较好的结果,但光学纹理特征在描述SAR图像纹理特性中因为相干成像特性往往失效.本文在前期工作纹理特征框架的基础上,提出了一种局部重要性采样二进编码的SAR图像纹理特征(Feature extraction based on local important sampling binary,LISBF)描述方法:首先,利用样本图像对局部采样位置进行随机自适应采样,基于重要性采样(Important sample,IS)方法输出递归学习位置结果;然后,利用学习出的纹理重要采样点对进行二进特征编码;最后,通过映射和统计生成描述算子.该特征较固定位置采样能够获取更大范围信息,同时能通过采样避免特征维数的急剧增大;通过自适应学习重要性关键点较随机采样更容易捕捉纹理固有信息;较好地适应了SAR图像极低信噪比和斑点现象的纹理.本文将该特征用于真实图像和标准纹理库的分类研究,实验结果证明了该特征的有效性.
[Abstract]:Synthetic aperture radar (SAR) images show a texture characteristic because of the speckle phenomenon and the spatial variation of the target response. The local image features such as local binary coding have better results in the optical texture description. However, in describing the texture characteristics of SAR images, the optical texture features often fail because of the coherent imaging characteristics. In this paper, a local importance sampling binary coding method for describing the texture feature extraction based on local important sampling binary SAR image is proposed. Firstly, random adaptive sampling of the local sampling location is performed using the sample image. Based on the importance sampling method, the recursive learning position is outputted. Then, binary feature coding is carried out by using the texture important sampling points. The description operator is generated by mapping and statistics. The feature can obtain a larger range of information than the fixed position sampling, and can avoid the sharp increase of feature dimension by sampling. It is easier to capture the inherent texture information by adaptive learning the key points of importance than random sampling, which adapts the texture of SAR images with extremely low SNR and speckle phenomena. In this paper, the feature is applied to the classification of real images and standard texture databases. The experimental results show the effectiveness of the feature.
【作者单位】：武汉大学电子信息学院信号处理实验室;
【基金】：国家重点基础研究发展计划(973计划)(2013CB733404) 国家自然科学基金(41371342,61331016) 中国博士后基金测绘遥感信息工程国家重点实验室专项科研经费资助~~
【分类号】：TN957.52

【参考文献】