北方农业学报 ›› 2020, Vol. 48 ›› Issue (1): 129-134.doi: 10.12190/j.issn.2096-1197.2020.01.25

• 农业经济·农业信息技术 • 上一篇    

基于随机森林法的农作物遥感识别方法研究——以阿荣旗部分区域为例

包珺玮, 于利峰, 乌兰吐雅, 许洪滔, 乌云德吉, 于伟卓   

  1. 内蒙古自治区农牧业科学院,内蒙古 呼和浩特 010031
  • 收稿日期:2019-11-12 出版日期:2020-01-19 发布日期:2020-05-08
  • 通讯作者: 乌兰吐雅(1970—),女,副研究员,硕士,主要从事遥感与GIS应用的研究工作。
  • 作者简介:包珺玮(1987-),男,助理研究员,硕士,主要从事资源开发与GIS应用的研究工作。
  • 基金资助:
    内蒙古农牧业科技创新基金项目(2018CXJJ09)

Research on crop remote sensing recognition method based on a random forest method —Take some areas of Arun Banner as an example

BAO Junwei, YU Lifeng, Wulantuya, XU Hongtao, Wuyundeji, YU Weizhuo   

  1. Inner Mongolia Academy of Agricultural and Animal Husbandry Sciences,Hohhot 010031,China
  • Received:2019-11-12 Online:2020-01-19 Published:2020-05-08

摘要: 【目的】为未来大尺度农作物遥感识别方法提供借鉴。【方法】 以内蒙古阿荣旗部分区域为例,采用Sentinel_2A影像主要数据源,利用随机森林法结合农作物光谱特征、纹理特征、参数特征,提取研究区的玉米、高粱、大豆、甜菜种植面积信息,并对分类结果进行精度验证和评价。【结果】 4种农作物的分类总体精度达到80.02%,Kappa系数为0.727 7,分类精度较好。【结论】 利用随机森林分类器结合均值飘移算法能够明显改善影像“椒盐现象”,并提高农作物的分类精度。

关键词: 随机森林法, 均值漂移, 农作物分类, 阿荣旗

Abstract: 【Objective】It can provide a reference for future large-scale crop remote sensing recognition method. 【Methods】We took the Sentinel_2A images as the main source of Arun Banner in Inner Mongolia. Based on a random forest method and the full utilization of the spectral features,texture features,and parameter features to extract the area of plants including corn,sorghum,soybean,and sugar,the classification results were analzyed. 【Results】The overall accuracy of the crop classification reached 80.02%,the Kappa coefficient achieved 0.727 7,and the classification results are good. 【Conclusion】Using random forest as a classifier combined with the way of mean shift it can ameliorate the phenomenon of “salt and pepper” and improve the accuracy of classification of crops.

Key words: Random forest method, Mean shift, Crop classification, Arun Banner

中图分类号: 

  • TP701