基于Landsat 8 OLI影像的山区园地提取方法

Extraction Method of Orchard in Mountainous Area Based on Landsat 8 OLI Image

我国园地种植面积在世界范围内位于前列,水果经济在现代农业经济中的作用比较明显。利用遥感技术对果园进行快速监测,准确掌握果园的面积与空间种植分布状况,对促进我国果园产业的可持续发展有重要的现实意义。目前对果园的提取大都集中在平原地区,对山区尤其是以云南为代表的云贵高原地区的果园提取比较少,对果园的提取方法也没有比较合适的。所以本文尝试采用随机森林这个比较常用的分类方法进行分类并且用决策树和支持向量机两种分类方法进行分类比较,提取以昭阳区为研究区域的园地。结果表明:在相同特征下与支持向量机和决策树分类方法相比随机森林分类结果更好;使用随机森林这个比较常见的分类算法的平均总体精度可以达到96%,分类算法的Kappa系数达到0.95,分类效果相对其他两种分类算法来说比较好。

The orchards planting area in China ranks the top across the board,and the role of fruit economy in modern agricultural economy is relatively obvious.Using remote sensing technology to quickly monitor the orchard and accurately grasp the area and spatial planting distribution of the orchard has great realistic meaning for facilitating the sustainable evolution of the orchard industry in China.At present,the orchard extraction is mostly concentrated in the plain area.There are few and no good extraction methods for orchards in mountainous areas,especially in Yunnan.Therefore,this article tries to use random forest,a more commonly used classification method,to assort and contrasts the sorts with decision tree and support vector machine,and extract the orchard with Zhaoyang District as the research area.The consequences demonstrate that the random forest classification consequences are better compared with the other two classification algorithms under the same characteristics.Average overall accuracy using stochastic forest classification algorithm is 96%and Kappa coefficient is 0.95.This classification effect is good.