基于热核共生矩阵的SAR图像纹理目标识别

SAR Image Target Recognition Based on Heat Kernel Co-Occurrence Matrix

相干斑噪声是合成孔径雷达(synthetic aperture radar,SAR)成像系统的固有缺陷,严重影响SAR图像的识别率。为此,提出了一种基于热核共生矩阵提取纹理特征统计量的算法,并将此方法用于SAR图像的纹理目标识别。首先通过构建图结构计算出图像每一点的热核特征,生成热核共生矩阵并计算纹理特征统计量;进而将热核特征与纹理特征统计量相结合构成特征矩阵;最后通过计算对应特征向量之间的相关系数,利用相似性度量的方法,实现SAR图像的纹理目标识别。实验结果表明,该方法可以识别SAR图像纹理目标,且识别效果要优于经典的基于灰度共生矩阵提取纹理统计量的方法。

Coherent spot noise is the inherent defect of SAR（ Synthetic Aperture Radar）,it seriously affect to the rate of SAR images recognition. In the process of SAR image recognition,we need select a characteristic without noise interference due to influence of the coherent spot. In this way,a novel method of extracting the texture feature was put forward and this feature was used in SAR image target recognition. Firstly,by building a graph structure,we computed the heat kernel feature at every point in SAR image. Furthermore,the heat kernel co-occurrence matrix was generated and its texture feature statistics were calculated; then we combined the heat kernel feature and texture feature statistics to form the characteristic matrix. Finally,we calculated the correlated coefficient of two SAR images and recognition was obtained by comparing similarities of the whole SAR images. This method,which is used to study the characteristics of heat kernel on graph,can allow full play to advantages of graph spectral theory. Experimental and their analysis show preliminarily that,compared with the method of classic gray co-occurrence matrix,this method,which is based on heat kernel co-occurrence matrix,shows higher recognition rate for SAR images.