摘要
利用BP神经网络和主成分分析法,结合SPOT-5遥感数据对土地开发整理区土壤有机质含量进行定量反演.试验采集了29个土壤样品并进行野外光谱测量,29个土样分为20个预测集和9个验证集,通过主成分分析对光谱信息进行变量转换,建立土壤有机质的BP神经网络预测模型,预测精度高达0.95.与土地开发整理前相比,土地开发整理后土壤有机质含量明显增加,土壤肥力提高而且分布均匀,土地平整效果显著,该方法对土地开发整理土壤质量验收工作具有重要的理论意义和实用价值.
Based on BP neural network and Principal Component Analysis(PCA)with SPOT-5 RS data, the soil organic matter (SOM) is quantitative retrievaled in the land development and consolidation region. 29 soil samples are collected from the tested region, 20 samples are the prediction sets, 9 samples are the test sets, the spectral information is transformed by Principal Component Analysis method, establishes the esti- mation models of SOM by BP neural network, the prediction accuracy is higher of 0.9S. The SOM content is higher after land consolidation, soil fertility is improved and distributed uniform, the effect of land consolida- tion is apparent,this method has important theoretical and experimental value in the soil quality inspection work of Land development and consolidation.
出处
《湘潭大学自然科学学报》
CAS
CSCD
北大核心
2012年第2期103-106,共4页
Natural Science Journal of Xiangtan University
基金
国家自然科学基金项目(41171397)
湖南省教育厅项目(10C0390)
长沙理工大学公路工程省部共建教育部重点实验室开放基金资助项目(kfj110102)
关键词
BP神经网络
土壤有机质
遥感定量反演
精度
BP neural network
soil organic matter
Remote Sensing quantitative retrieval
accuracy