产量数据误差的识别与过滤方法研究  

Research on identification and filtering method of yield data error

作　　者：马鑫 王鹤[2] 李世国[2] 金学伟[2] 白晓平[3] 

机构地区：[1]辽宁福鞍重工股份有限公司,辽宁鞍山114016 [2]辽宁科技大学,辽宁鞍山114051 [3]中国科学院沈阳自动化研究所,沈阳市110016

出　　处：《中国农机化学报》2016年第11期139-145,共7页Journal of Chinese Agricultural Mechanization

基　　金：国家高技术研究发展计划(863计划)资助项目(2013AA040403);辽宁省科技攻关计划资助项目(Y5L7160701)

摘　　要：针对产量数据包含大量误差的问题,首先研究分析产量数据的获取过程,从统计分析的角度,将可能造成产量数据误差的归纳为系统误差、随机误差和粗大误差。然后分别针对不同产量数据的误差来源,提出误差数据处理方法,设计并实现6种过滤器:割台状态、填充和排空延时、传感器数据异常、不切实际的产量值、局部空间的异常值、短收割段和坐标重叠。最后仿真实验表明,过滤后产量数据由随机因素引起的产量数据空间变异得到显著改善。Systematic errors,random errors and gross errors existing in the yield data acquisition process were analyzed and classified based on statistic analysis.Different error data processing methods were proposed aimed at different error sources.Six kinds of filter were designed and put forward,including combine header status up,filling and emptying time,sensor outliers,values exceeding biological yield limits,local neighborhood outliers,short segments,and co-located points.The simulation experiments indicated that spatial variability caused by random factors has been significantly improved after using the yield data filtering method.

关 键 词：精准农业 产量误差 误差分析 误差识别与处理 

分 类 号：S24[农业科学—农业电气化与自动化] TP29[农业科学—农业工程]