基于深度神经网络的粮仓储粮数量检测模型

Granary storage quantity detection model based on deep neural network

[目的]为满足国家对全国储粮数量在线检测的迫切需求,提出了一种基于深度神经网络的粮仓储粮数量检测方法。[方法]通过在粮仓底部布置若干压力传感器的方法获取粮仓底部所受压强值,并以分次进粮方法,分别记录所受压强值。通过R语言平台构建不同层次的深度神经网络结构并利用对数据集的学习得出检测模型,根据检测精度选择出最佳检测模型结构。通过最佳检测模型分别对试验仓及通州、齐河实仓进行检测实验。[结果]试验仓检测平均误差约为1.88%,通州实仓检测平均误差约为0.02%,齐河实仓检测平均误差约为0.08%。[结论]基于深度神经网络的粮仓储粮数量检测模型精度高,可用性强,为粮仓储粮数量的检测提供了一种新方法。

[Objectives]In order to meet the urgent need of national on-line detection of grain storage quantity,a method of grain storage quantity detection based on depth neural network was proposed.[Methods]The pressure values at the bottom of the granary were obtained by arranging some pressure sensors at the bottom of the granary,and the pressure values were recorded by grading feeding method.The R language platform was used to construct different layers of deep neural network structure,and the detection model was obtained by learning the data set.According to the detection accuracy,the best detection model structure was selected.Through the best detection model,the experimental warehouse and Tongzhou and Qihe real warehouse were tested.[Results]The average error of the experimental warehouse was about 1.88%,that of the Tongzhou warehouse was about 0.02%,and that of the Qihe warehouse was about 0.08%.[Conclusions]The model based on depth neural network has high accuracy and high availability,which provides a new method for the detection of grain storage.