Design and experimental research on dielectric sensor for detection of available calcium content in paddy soil

<bold>Objective</bold> The scale of paddy soil in China is large.Calcium is an important element required for the growth of rice and also an important indicator for measuring the fertility of paddy soil.In order to solve the problem of inaccurate application of calcium fertilizer which is necessary for rice growth，rapid detection on the available calcium content that can be directly absorbed by crops such as rice was conducted. <bold>Method</bold> In this study，a dielectric sensor for detecting the available calcium content in paddy field soil was designed by combining dielectric properties with LoRa technology for the actual environment of paddy fields.Based on the test results，dielectric detection models for the available calcium content in paddy field soil were established.The sensor designed in this study is mainly divided into four parts：the upper cover of the sensor，the probe structure，the internal circuit and the probe.The hardware circuit adopted the STM32 single-chip microcomputer as the main control to control the sweep frequency signal generation module centered on AD9859 for sweep frequency detection.The detected signals were converted by the AD module to obtain the amplitude ratio and phase difference data.In addition，a suitable power module had been designed for power supply，and a LoRa module was used for wireless communication.Based on the hardware design，the programs for the two stages of data acquisition and device application were written respectively，mainly including the initialization program，the AD9859 frequency scanning program，the key program，the data conversion program，the prediction algorithm program for soil available calcium content，the LoRa configuration program and the serial port program.In this study，the Competitive Adaptive Reweighted Sampling（CARS method was used to extract characteristic parameters from five types of data—amplitude ratio，phase difference，amplitude ratio voltage，phase difference voltage，and the combination of amplitude ratio voltage and phase difference voltage—of soil samples with different available calcium contents under moisture contents of 20%，25%，30%，35%，and 40%，respectively.Corresponding Support Vector Regression（SVR）prediction models were also established. <bold>Result</bold> SVR prediction models were established by using phase difference voltage，phase difference voltage，amplitude ratio，phase difference，amplitude ratio voltage combined with phase difference voltage as inputs at moisture contents of 20%，25%，30%，35% and 40% respectively，and were written into the sensor firmware for the prediction of available calcium content in soil.The test results showed that the <italic>R²</italic> of the prediction models under the five moisture contents were 0.9856，0.9911，0.9906，0.9915 and 0.9918 respectively，all above 0.98 with high accuracy.The <italic>RMSE</italic> range was 69.008-83.647 mg/kg.By extracting the features of the frequency points，the running speed of the model had been significantly improved，making a single detection only take 2 minutes. <bold>Conclusion</bold> This sensor is portable and practical，with high accuracy and fast speed in detecting the available calcium content in paddy field soil.Meanwhile，the sensor has all-weather working capability and can achieve remote online real-time monitoring in combination with LoRa Internet of Things.