一种快速的SAR图像分类方法

SAR图像分类是＆讯图解译中非常重要的环节，但由于SAR图像中相干斑噪声的存在，使得传统方法不能很好地对SAR图像进行分类。再之，SAR图像分类具有计算量大、耗时长的特点，SAR所能获取的信息数据量也越来越大，如何快速、准确地对SAR图像进行分类以及时获取有用信息显得日益迫切。本文提出了一种快速的SAR图像分类方法，该方法将图像的空间域和频域特征相结合，并基于并行计算环境，对图像中的每一点都计算相应的小波能量特征、共生灰度矩阵特征和滤波后的灰度特征，并组成特征向量对SAR图像进行分类。实验结果证明该方法能取得较好的分类效果，且速度较快。

A Fast Method of SAR Image Classification

SAR image classification is a key step in SAR image interpretation,but for the existing of speckle in SAR image,SAR image can't be classified well with traditional methods.Furthermore,it usually takes long time to classify SAR image due to vast calculation,and the data obtained by SAR is larger and larger,how to classify SAR image precisely and quickly appears more and more urgent.Based on the feature of SAR image and the advantage of parallel computing in high performance computing,a fast method of parallel SAR image classification is proposed.The features in spatial domain and in frequency domain are combined,features in the both domains are calculated in the parallel computing environment,and finally SAR image is classified according to all feature vectors recalled from all the processes.The experiment results show that good result of classification can be obtained quickly with the method.