摘要
耕地是农业发展的基础,而遥感技术是当前监测耕地面积以及分布情况的主要工具。耕地本身复杂的地形特征以及其他背景地物的混淆,使得地形特征、分辨率、空间物理误差、几何变形、算法等一直是制约耕地快速实时监测的主要因素。SVM算法具有小样本学习、抗噪声性能好、学习效率高、鲁棒性好等优点。通过SVM算法对江苏省某地级市的卫星遥感图像分类,识别出其中的耕地并划分,对耕地的分类准确率达到了90%以上。实验结果表明,使用SVM算法能够快速高效地对遥感图像中的耕地进行识别并划分。
Cultivated land is the basis for developing the agriculture. Remote sensing technology is the main tool for monitoring the current situation and distribution area of cultivated land. However,the complicated terrain features and the confusion of other background features make the topographic features,resolution,spatial and physical errors,geometric deformations,and algorithms restrict the speed of real-time monitoring of cultivated land all the time. The Support Vector Machines( SVM) algorithm has the advantages of small sample learning,good noise immunity,high learning efficiency and good robustness. In this paper,we classify the satellite remote sensing images of Yizheng City,Yangzhou City,Jiangsu Province by SVM algorithm and identify the cultivated land and divide them. Classification accuracy rate reached 90%. Experimental results show that the SVM algorithm can quickly and efficiently identify and classify cultivated land in remote sensing images.
作者
李昌俊
黄河
李伟
LI Chang-jun;HUANG He;LI Wei(Institute of Intelligent Machines,Chinese Academy of Sciences,Hefei 230031,China;University of Seienee and Technology of China,Hefei 230027,China)
出处
《仪表技术》
2018年第11期5-8,48,共5页
Instrumentation Technology
基金
国家自然科学基金面上项目(31671586)
关键词
遥感图像分类
耕地划分
支持向量机
remote sensing image
cultivated land division
support vector machines (SVM)
作者简介
李昌俊(1993-),男,在读硕士研究生,主要从事农业大数据、深度学习方面的研究.
