CSI-Based Rapid Moisture Detection Technology And System For Grain Storage

Food security is of paramount importance in ensuring national security,improper grain storage has led to annual grain losses in China of hundreds of thousands of tons.One of the main reasons for this is a delay in obtaining accurate grain information,resulting in missed opportunities for optimal drying.To prevent such losses,it is essential to have comprehensive and timely access to grain information.Grain moisture is a vital indicator of grain condition,and traditional moisture testing technology suffers from the drawbacks of long testing times and low accuracy.While modern technology can provide accurate results,the high cost of equipment limits its widespread use.Thus,there is an urgent need for a low-cost,fast,and accurate grain moisture detection method,which wireless sensing technology can satisfy.In this study,we investigate a rapid grain moisture detection technology based on Wi Fi signals,taking advantage of the fact that grains with different moisture levels will have different effects on Wi Fi signal propagation.We use the ESP32 Wi Fi chip for channel state information(CSI)acquisition,reducing the complexity of acquiring CSI devices.During the data analysis stage,two feature selection algorithms,namely RF(random forest)and PCA(principal component analysis),were employed to extract feature subcarriers.These were then combined with two classification algorithms,namely CNN(convolutional neural network)and BLS(broad learning system),to evaluate the pros and cons of each method.The optimal model for detecting grain moisture was obtained using PCA-BLS.The optimal parameters of the BLS algorithm were obtained by combining it with an improved BLS algorithm and changing the number of feature nodes and enhancement nodes.The effect of changing the distance between antennas on the accuracy rate was also studied.Subsequently,a grain moisture recognition device was manufactured,and a mobile platform was selected as the computing terminal.A visualization interface was designed to reduce the learning cost of using the device.The performance of the prototype device was tested,including the detection of standby current and power consumption at full load.Although the analysis time was extended after transferring the data analysis to the mobile platform,it was still within an acceptable range overall.In summary,the main innovation of this paper is to apply a low-cost CSI collection device for grain moisture detection and to obtain an optimal model for grain moisture identification through experiments.A prototype device for grain moisture detection is designed,which has the features of high detection accuracy,easy deployment,fast detection speed and low cost,and has a broad development prospect in the future.