基于K-均值聚类的绿洲农田管理分区提取的研究  

Delineation of Precision Agriculture Management Zones in Oasis Field Based on K-means Algorithm

作　　者：安光辉[1] 马蓉[1] 陈伟[1] 芦帅[1] 

机构地区：[1]石河子大学机械电气工程学院,石河子832003

出　　处：《石河子大学学报（自然科学版）》2011年第6期757-761,共5页Journal of Shihezi University(Natural Science)

基　　金：国家"863"高技术研究发展计划项目子课题(2010AA101403);教育部科学技术研究重点项目(209144);石河子大学高层次人才科研启动资金专项(ZRKX200735);石河子大学"263"人才培养工程项目

摘　　要：根据产量数据的空间变异划分管理分区是确定精准农业变量作业单元的有效途径,可以为样区土壤管理提供科学的决策依据。本研究以新疆绿洲农田为研究对象,通过安装在John Deer 9970采棉机上的Insight产量监测系统获取棉花产量实时数据,并以之为数据源进行农田精准管理分区的研究。研究采用K-均值聚类算法提取管理分区,并以方差减少率的变化来评价管理分区划分的效果。结果表明:当过滤窗口尺度为18m×18m,管理分区数为4时,各分区内的棉花产量变异较小且分区间棉花产量变异较大,分区尺度适合变量作业。Zoning based on spatial variability of yield on a field scale is a key to variable rate application in precision agriculture.The objective of this research was to delineate precision agriculture management zones of oasis cotton field.Firstly by means of the Insight yield monitor system installed on the JD 9970 cotton picker,the real time cotton yield data is obtained.Then K-means clustering algorithm was used to delineate management zones with the yield data as main data source.In order to determine the optimum management zones,the relative variance reduction was used to judge its reasonability.The results showed that when the size of the filter window is 18m×18m and the whole field is divided into four management zones the regional differences all decreased obviously and the difference between regions is very significant.Moreover,the region scale of the management zones delineated with this method is suitable for variable-rate application.The delineated management zones can be used for fertilizer recommendation to manage soil nutrient and salinity more efficiently in coastal saline field.

关 键 词：K-均值聚类 管理分区 精准农业 绿洲农田 

分 类 号：S231[农业科学—农业机械化工程] S126[农业科学—农业工程]