基于多特征和WSVM的SAR图像河流目标检测

Target detection algorithm for rivers in SAR images based on multi-features and WSVM

为进一步提高合成孔径雷达(synthetic aperture radar,SAR)图像中河流目标检测的准确性,本文提出了基于多特征和小波支持向量机(wavelet support vector machine,WSVM)的SAR图像河流目标检测方法。首先使用均值比表示像素点邻域的灰度特征,Gabor小波提取其纹理特征,并将其融合构造训练样本;然后将归一化处理后的特征矩阵输入WSVM进行训练,并利用训练好的WSVM对图像的每个像素点进行分类;最后根据河流的区域连通性和面积、形状特征,去除阴影、湖泊等与河流相似的区域。大量实验结果表明,与其他河流目标检测方法相比,本文方法检测的河流目标更加完整,背景与河流的误分区域更少,河流边缘保持得更好。

In order to further improve the accuracy of river target detection in synthetic aperture radar （SAR） images, a method of river target detection based on multi-features and wavelet support vector machine （WSVM） in SAR images is proposed. Firstly, gray features of pixel neighborhood are represented by the mean ratio. Texture features are extracted by Gabor wavelet. The training samples are constructed by fusion of the extracted gray features and texture features. Then, the normalized feature matrix is inputted into the WSVM for training. Each pixel in the images is classified by the trained WSVM. Finally, the similar regionals with rivers such as shadows, lakes are removed according to the regional connectivity, areas and shape features of rivers. A large number of experimental results show that compared with other methods of river target detection, the proposed method has more completely detection, error regions of classification are much less and edges of rivers are preserved better.