首页|基于D-S证据理论的农作物气候品质预测方法研究:以晚熟杂交柑橘春见为例

基于D-S证据理论的农作物气候品质预测方法研究:以晚熟杂交柑橘春见为例

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
[目的]基于多源气象数据构建果实品质(糖含量等级)预测模型,为科学评价果实气候品质及深入挖掘农产品气候资源提供科学依据.[方法]以晚熟柑橘春见果实为研究对象,利用多源数据融合技术、人工神经网络(BP神经网络、RBF神经网络和Elman神经网络)和D-S证据理论,包括气象数据质量控制、特征选取、特征级融合、决策级融合 4个步骤,构建基于多源气象数据的果实品质(糖含量等级)预测模型.[结果]春见果实品质预测模型采用BP神经网络预测结果总体准确率为 87.50%,平均绝对误差(MAE)为 0.150,均方根误差(RMSE)为 0.447;RBF神经网络预测结果总体准确率为 85.00%,MAE为0.175,RMSE为 0.474;Elman神经网络预测结果总体准确率为 87.50%,MAE为 0.150,RMSE为 0.447;D-S证据理论决策融合总体预测准确率达 95.20%,分别较BP神经网络、RBF神经网络和Elman神经网络提升 7.7百分点、10.2百分点和 7.7百分点,MAE和RMSE分别为 0.040和 0.214,均明显降低.[结论]D-S证据理论决策融合后的果实品质预测准确率相比单一神经网络预测更高、误差更小.
Research on Prediction Method for Crop Climate Quality Based on D-S Evidence Theory:Taking late mature hybrid citrus Harumi as an example
[Objective]A model for predicting the fruit quality(sugar content level)based on multi-source meteorological data was established,which provided the scientific basis for evaluating the climate quality of fruit and deep exploiting climatic resources of agricultural product.[Method]Based on multi-source data fusion technology,artificial neural networks(BP neural network,RBF neural network and Elman neural network)and D-S evidence theory,including meteorological data quality control,feature selection,feature level fusion and decision level fusion,a model for predicting the fruit quality(sugar content level)based on multi-source meteorological data was constructed by taking late mature citrus Harumi as the research object.[Result]The overall accuracy of prediction results for Harumi fruit quality using BP neural network was 87.50%,the mean absolute error(MAE)was 0.150 and the root mean square error(RMSE)was 0.447.The overall accuracy of prediction results using RBF neural network was 85.00%,with MAE of 0.175 and RMSE of 0.474.The overall accuracy of prediction results by Elman neural network was 87.50%,with MAE of 0.150 and RMSE of 0.447.The overall accuracy of prediction results by D-S evidence theory decision fusion reached 95.20%,which was improved by 7.7 percent points,10.2 percent points and 7.7 percent points compared to BP neural network,RBF neural network and Elman neural network,respectively.The MAE and RMSE were respectively 0.040 and 0.214,which reduced significantly.[Conclusion]The accuracy of predicting the fruit quality after D-S evidence theory decision fusion is higher and the error is smaller compared to single neural network prediction.

late mature citrusHarumiclimatic qualitymulti-source data fusionBP neural networkRBF neural networkElman neural networkD-S evidence theory

付世军、李梦、杨晓兵、何震、袁佳阳、刘书慧、徐越、卢德全、张利平

展开 >

南充市气象局,四川 南充 637000

南充市果树技术指导站,四川 南充 637000

达州市气象局,四川 达州 635000

晚熟柑橘 春见 气候品质 多源数据融合 BP神经网络 RBF神经网络 Elman神经网络 D-S证据理论

四川省科技厅农业重点研发项目南充市科技局推进乡村振兴科技创新项目

2023YFN005122XCZX0005

2024

10.3969/j.issn.1001-3601.2024.05.017

贵州农业科学
贵州省农业科学院

贵州农业科学

CSTPCD
影响因子：0.642
ISSN：1001-3601
年,卷(期)：2024.52(5)
  • 20