论文部分内容阅读
针对由实际遥感地物类型难以确定导致的多光谱遥感影像变化检测精度较低的问题,提出一种基于SVM混合核的遥感图像变化检测。首先利用CVA算法构造差异影像,其次利用灰度共生矩阵提取差异影像的纹理特征与差异影像的灰度特征组成特征向量,接着利用差异影像的直方图选择置信度高的训练样本,并利用构造的SVM混合核进行训练得到分类超平面,最后利用SVM混合核函数对差异影像进行二分类得到最后的变化检测结果。实际遥感数据验证结果表明,所构造的SVM混合核函数用于多光谱遥感影像变化检测中是可行、有效的。
Aiming at the problem of low detection precision of multispectral remote sensing image caused by the difficulty in determining the type of remote sensing object, a remote sensing image change detection based on SVM hybrid kernel is proposed. Firstly, the CVA algorithm is used to construct the difference image. Secondly, the gray-level co-occurrence matrix is used to extract the texture feature of the difference image and the gray feature of the difference image to form the eigenvector. Then the histogram of the difference image is used to select the training samples with high confidence. SVM mixed kernel to get the classification hyperplane. Finally, the SVM mixed kernel function is used to classify the difference images to get the final change detection result. The results of actual remote sensing data show that the proposed SVM mixed kernel function is feasible and effective for the detection of multi-spectral remote sensing image changes.