Soil total nitrogen (TN) concentration is an important soil nutrient parameter. However, the high-precision measurement of TN concentration in real time remains a challenge. Accordingly, an on-the-go TN concentration detector was developed on the basis of near-infrared spectroscopy. The modular concept and the extreme learning machine (ELM) algorithm were applied to design and model the detector, respectively. The detector consisted of three modules: a mechanical unit used to support the entire detection system, an optical unit for sensing soil in the soil detection trench and a control unit including the embedded ELM model for obtaining real-time readings of TN concentration. The ELM model was established using absorbance data at 1680, 1550, 1375, 1245, 1130 and 1070 nm. Laboratory calibration and validation experiments were conducted to verify the performance of the detector. The coefficient of determination (R²) of the TN concentration of the ELM estimation model was 0.90, and the validation R² was 0.82. Laboratory results showed that the developed detector exhibited stable performance and high precision. Field tests were conducted and the correlation coefficients between standard laboratory measurements and the detector were 0.82 in field operation and 0.91 in the laboratory. The experimental results showed that the detector can quantify TN in the field in real time with stable performance and high accuracy.

土壤总氮（TN）浓度是重要的土壤养分参数。但是，实时高精度测量TN浓度仍然是一个挑战。因此，基于近红外光谱技术开发了一种移动式TN浓度检测器。模块化概念和极限学习机（ELM）算法分别用于设计和建模探测器。该探测器由三个模块组成：一个用于支撑整个探测系统的机械单元，一个用于感测土壤探测沟中土壤的光学单元以及一个包含嵌入式ELM模型的控制单元，用于实时获取TN浓度读数。使用1680、1550、1375、1245、1130和1070 nm的吸光度数据建立ELM模型。进行了实验室校准和验证实验，以验证检测器的性能。确定系数（ELM估计模型的TN浓度的R ²）为0.90，验证的R ²为0.82。实验室结果表明，研制的检测器性能稳定，精度高。进行了现场测试，标准实验室测量值与检测器之间的相关系数在现场操作中为0.82，在实验室中为0.91。实验结果表明，该检测器能够实时，定量，高精度地定量田间总氮。