Rapid Detection Of Stored Grain Moisture Based On Commercial WIFI Signal

China is a country with a huge population.Grain security is linked to the national economy,people's livelihood and national development.The period of grain storage in our country is three to five years.In this process,we are required to detect the temperature,moisture,pest,mildew and other multiple indicators to ensure grain security.One of the important indicators of grain security is the moisture content.When the moisture content is too high,it will lead to problems such as mildew and deterioration of grain.On the contrary,when the moisture content is too low,the organic matter inside the grain will be damaged,thus reducing the quality of grain.At the same time,the moisture content of grain is also associated with the interests of the sellers and consumers in the following links of grain processing and sales.Therefore,detection of grain moisture in grain storage is a key point to ensure grain security.However,the traditional methods of grain moisture detection have some problems,such as time-consuming,low efficiency,damaging to grain itself,unable to detect on-line and so on.And there are some limitations in modern methods of grain moisture detection,such as low precision and expensive instrument.Based on the above problems,this paper proposes a method to detect grain moisture using commercial WIFI equipment.The research work of this paper includes:Firstly,based on the in-depth understanding of the background of grain moisture detection technology and the principle of WIFI wireless communication technology,this paper analyzes the grain moisture and WIFI wireless communication technology,and provides the feasibility of using Channel State Information(CSI)in WIFI wireless communication technology to detect grain moisture.It can detect grain moisture by the influence of WIFI signal on the amplitude and phase information of CSI in the process of different grain moisture transmission.Secondly,this paper designs a grain moisture abnormality detection system based on Support Vector Machine(SVM).CSI data of normal and abnormal moisture wheat samples is collected by the commercial WIFI equipment.Then preprocessing the collected CSI data which including outlier elimination,noise elimination,feature extraction and normalization.Finally,the abnormal detection system of grain moisture implemented with SVM algorithm model.The performance of the system is verified in the scene of sight distance and non-sight distance.The results show that the average accuracy of the system detection for wheat samples from normal and abnormal moisture is above 90%.Finally,the impact of system performance is analyzed.Finally,due to the limitations of traditional machine learning methods,this paper designs a precise grain moisture detection system based on Long Short-Term Memory(LSTM).Through the commercial WIFI equipment,CSI data of 10 kinds of moisture wheat samples with 8% ~17% moisture content is collected,and the system moisture label is calibrated and then the collected CSI data is reprocessed.The double-layer LSTM network and Softmax classifier are designed for off-line training and on-line testing,and the grain moisture accurate detection system is realized.The performance of the system is verified in the scene of sight distance and non-sight distance.The results show that the average accuracy of the system detection for wheat samples of ten kinds of moisture is above 95%.Finally,the influence of system parameters and different experimental scenarios on system performance is analyzed.The main innovation of this paper lies in the first using of commercial WIFI equipment and related technologies to build a grain moisture detection system.The system gets the advantages of easy deployment,fast detection speed and low cost.And it makes a very broad research and development prospects.