基于BP神经网络的土壤养分空间插值

以广东省增城市为研究对象，采集全市内200个土壤样点，利用BP神经网络插值方法对研究区土壤的氮和磷进行空间插值预测，将插值结果与土壤样点实测值进行对比，得到预测数据的误差均方根。结果表明，BP神经网络的插值精度比克里格高，在样点较少的情况下．BP神经网络的插值结果克服了克里格插值方法的平滑效应。BP神经网络对插值的样本数据的分布类型没有要求，比传统插值方法有更强的泛化能力，是一种可替代的插值方法。...

Full description

Saved in:

Bibliographic Details
Published in	广东农业科学 Vol. 40; no. 7; pp. 64 - 67
Main Author	程家昌黄鹏熊昌盛胡月明张真廖琪
Format	Journal Article
Language	Chinese
Published	华南农业大学信息学院,广东广州,510642%华南农业大学信息学院,广东广州510642 2013 广东省土地利用与整治重点实验室,广东广州510642
Subjects	BP神经网络克里格插值土壤养分空间预测土壤养分空间预测 BP神经网络克里格插值
Online Access	Get full text
ISSN	1004-874X

Cover

More Information
Summary:	以广东省增城市为研究对象，采集全市内200个土壤样点，利用BP神经网络插值方法对研究区土壤的氮和磷进行空间插值预测，将插值结果与土壤样点实测值进行对比，得到预测数据的误差均方根。结果表明，BP神经网络的插值精度比克里格高，在样点较少的情况下．BP神经网络的插值结果克服了克里格插值方法的平滑效应。BP神经网络对插值的样本数据的分布类型没有要求，比传统插值方法有更强的泛化能力，是一种可替代的插值方法。
Bibliography:	44-1267/S 200 soil sampling points were collected in Zengcheng city which was the research object in Guangdong. We designed two kinds of layout programs to study the method used in the spatial prediction of soil properties： Kriging method and BP neural network method. And then analysed the forecast error in the test set. We found that the interpolation accuracy of the BP neural network in the same circumstances was higher than Kriging method. BP neural network method did not require the distribution of sample data. In the case of fewer sample points, the results of the BP neural network method was better what showing the advantages of an alternative interpolation method. BP neural network; soil properties; spatial prediction; Kriging CHENG Jia-chang, HUANG Peng, XIONG Chang-sheng, HU Yue-ming ZHIANG Zhen, LIAO Qi （1.College of lnformatics, South China Agricultural University, Guangzhou 510642,＇ China,＂ 2.Key Laboratory of Guangdong Province Land Use and Remediation, Guangzhou 510642, China）
ISSN:	1004-874X