【摘要】 [目的]有效磷是影响水稻生长和产量的关键土壤养分指标。传统实验室检测方法耗时长、成本高，难以满足快速监测需求。为此，设计一种稻田土壤有效磷在线检测传感器，以实现土壤养分快速监测，为水稻科学施肥提供支持。[方法]以STM32F405RGT6微控制器为核心平台，集成了信号采集模块与低功耗无线传输模块。系统软件基于MDK 5.0开发环境，采用C语言编写，实现信号生成、数据转换、通信以及关键的插值算法等核心功能。通过采集到的传感器响应数据，系统应用CARS特征波长选择算法，有效提取幅值比、相位差以及融合幅值比与相位差的数据特征。利用提取的特征数据，构建支持向量回归（SVR）预测模型，最终实现对土壤有效磷含量的快速检测。[结果]所建模型（R²≥0.94）能有效预测稻田土壤有效磷含量，但预测精度受含水率影响。综合对比后，优选各含水率下的最佳模型。在20%、25%、35%、40%含水率下采用幅值比模型，其R²为0.979,0.954,0.991,0.978,RMSEP为7.78,11.26,5.44,7.82 mg·kg^-1；在30%含水率下，采用幅值比+相位差模型，R²为0.988,RMSEP为5.67 mg·kg^-1。[结论]该装置对于稻田土壤有效磷含量具有较高精度的检测，同时设备具有全天候工作能力，配合LoRa物联网可实现远程在线实时监测。更多还原

【Abstract】 [Objective]Available phosphorus is a key soil nutrient indicator affecting rice growth and yield. Traditional laboratory testing methods are time-consuming and costly, making it difficult to meet the demand for rapid monitoring. Therefore, this study designed an online sensor for detecting available phosphorus in paddy soil to achieve rapid soil nutrient monitoring and support scientific fertilization for rice. [Methods]The proposed system employs a high-performance STM32 F405 RGT6 microcontroller as its core processing platform, integrating a signal acquisition module and a low-power wireless transmission module. The system software,developed within the MDK 5.0 environment using the C programming language, implements core functionalities including signal generation, data conversion, communication protocols, and a critical interpolation algorithm for data preprocessing. Utilizing the acquired sensor response data, the system applies the CARS algorithm for feature wavelength selection. This enables the effective extraction of key data features: amplitude ratio, phase difference, and fused features combining amplitude ratio with phase difference.Subsequently, these extracted features are utilized to construct a SVR prediction model. Ultimately, this integrated approach achieves rapid detection of soil available phosphorus content. [Results]The experiments showed that the established model（R² ≥ 0.94） could effectively predict the available phosphorus content in paddy soil, but the prediction accuracy was affected by the moisture content.After comprehensive comparison, the best models for each moisture content were selected. The amplitude ratio model is used at moisture contents of 20%, 25%, 35%, and 40%, the R² values are 0.979, 0.954, 0.991, and 0.978, and RMSEP values are 7.78,11.26, 5.44, and 7.82 mg·kg^-1, respectively. At a moisture content of 30%, the amplitude ratio + phase difference model is used, the R² value is 0.988 and the RMSEP value of 5.67 mg · kg^-1. [Conclusion]This device has a high-precision detection capability for available phosphorus content in paddy soil. Moreover, the equipment has all-weather working capability and, in combination with the LoRa Internet of Things, can achieve remote online real-time monitoring.更多还原