论文部分内容阅读
基于国产资源一号02C星高分辨率(high resolution,HR)影像,提取了基于变差函数、灰度共生矩阵和梯度的多纹理特征,结合光谱信息构建了基于支持向量机(support vector machine,SVM)的多源信息复合模型的图像分类方法,并与传统最大似然法和决策树法的分类结果进行了比较。研究表明,变差函数纹理和梯度纹理参与的多源复合数据有效提高了图像的分类精度,总分类精度由85.14%提高到87.43%,Kappa系数由0.82提高到0.85;绝对值变差函数为纹理最佳窗口分析提供了理论依据,基于累积步长提取的纹理特征能显著提升图像分类的精度,分类准确率提高了13.94%,Kappa系数增加了0.17;基于多源复合数据的SVM高空间分辨率遥感图像分类方法能有效解决传统图像分类结果破碎的问题,比最大似然方法和决策树法的分类精度显著提高,总精度达到89.14%,Kappa系数为0.87,分别提高了6.85%和10.84%。实验表明,ZY-1 02C星HR数据在冬小麦信息提取中具有一定的稳定性和优势。
Based on the high resolution (HR) images of domestic No.1 02C, the multi-texture features based on variogram, grayscale co-occurrence matrix and gradient were extracted and combined with the spectral information to construct a support vector machine based on support vector machine SVM) multi-source information composite model of image classification method, and the traditional maximum likelihood and decision tree classification results were compared. The results show that the multi-source composite data with variogram texture and gradient texture effectively improve the classification accuracy of the image, the total classification accuracy is increased from 85.14% to 87.43%, and the Kappa coefficient is increased from 0.82 to 0.85. The absolute value variation function is the texture The optimal window analysis provides a theoretical basis. The texture feature extracted based on the cumulative step size can significantly improve the accuracy of image classification. The accuracy of classification is improved by 13.94% and the Kappa coefficient is increased by 0.17. Based on multi-source composite data, SVM high spatial resolution Compared with the maximum likelihood method and the decision tree method, the classification accuracy of the remote sensing image classification method can effectively solve the problem of the traditional image classification. The total accuracy is 89.14% and the Kappa coefficient is 0.87, which increases by 6.85% and 10.84% respectively. Experiments show that the ZY-1 02C star HR data has certain stability and advantage in the winter wheat information extraction.