基于纹理分析的航空影像耕地信息提取

摘要

In the face of the situation that the cultivated land is being decreased and the population - land contradiction is becoming very obvious,the cultivated land information in the remote sensing image was extracted accurately and had fast become the basis of cultivated land dynamic monitoring. In this paper,the aerial image of 0. 5m resolution in a county of Mianyang City was chosen as the experimental data. The textures of the aerial im-age were firstly analyzed by using Gray Level Co - occurrence Matrices. Then the cultivated land information was extracted by the maximum likelihood classification. The results show that:the overall classification precision is more than 90% ,and the coefficient of kappa is more than 0. 81.%面对耕地面积日益锐减，耕地矛盾日益突出的形势，准确快速的提取遥感影像中的耕地信息已成为耕地动态监测的基础.本研究以绵阳市某县0.5 m 分辨率的航空影像为实验数据，借助 RS 和 GIS 技术，采用共生矩阵分析航空影像纹理特征，在此纹理影像上运用最大似然监督分类方法对耕地信息进行了自动提取.实验结果表明：耕地信息提取正确率大于90％，kappa 系数大于0.81.该方法提取速度快，自动化程度高，精度好，能够有效利用地物的空间分布信息和结构信息.