结合纹理特征的SVM岛礁信息提取分析

海岛生态系统具有丰富的生物多样性和重要的生态功能,近年来,我国海岛的生态环境正遭受人类活动的严重破坏,目前,需要通过海岛调查来掌握海岛利用信息,为保护和修复海岛生态环境提供技术支撑。运用高空间分辨率影像,基于珊瑚礁岛的遥感特性,耦合纹理特征,借助支持向量机良好的分类性能,特别在小样本、非线性及高维特征空间具有较好的泛化能力,构造支持向量机理论与算法。针对珊瑚礁岛信息提取,对SVM的核函数进行比较,可获得较为理想的函数,融合纹理变化参数能够较好地区分礁坪与沙滩,能够有效提取建筑用地信息。在此基础上,运用SVM方法再次进行分类,对图像进行预处理可提高分类精度,精度可达77.8%。

Analysis Coupling Features with SVM to Extract Information of Island Land Use

Island has high biodiversity and important ecological function.In recent years,China’s ecological environment of the islands are suffering from serious damage by human activities.Nowadays,we need investigate the island features to provide technical support for the island ecological environment protection.By using high spatial resolution images,based on coral reefs island remote sensing properties,with the aid of the Support Vector Machine(SVM)which has good classification performance,especially with small sample,nonlinear and high dimensional feature space,has good generalization ability to build structural support vector machine(SVM)theory and algorithm.According to the coral information extraction,comparing the kernel function of SVM to obtain ideal function,image preprocessing can improve the classification accuracy.