Abstract: Nitrate is an important source of plant nitrogen nutrition, and low-cost, minimally destructive in situ detection of nitratenitrogen content in deep soil has been a bottleneck. This study designed a low-cost, deep soil nitrate nitrogen in situ real-time detection system based on soil spectral characteristics. The system is based on a single spectral band, which can be deeply implanted into the soil for in situ measurement. A titanium sintered cartridge was used to collect the soil solution, and a 220 nm band UV lamp was used as the irradiation light source to detect the nitrate-nitrogen content in the collected soil solution according to the principle of absorbance and the voltage-current conversion and amplification circuit, and the data were also autonomously transmitted to the supercomputer using the LoRa wireless transmission module, with no need of human intervention, no need to collect and process the soil samples, which not only greatly reduced the detection cost, but also reduced the labor cost. The prediction coefficient of determination R~2, correlation coefficient R, and root mean square error RMSE of the soil nitrate nitrogen prediction model were 0.982 2,0.991 1 and 0.009 2, respectively, which satisfied the accuracy of the measurement for practical applications.