In the near infrared spectral region, a reflection plant water detection instrument was designed by using microcontroller STC12C5A60S2. The instrument consisted of signal acquisition system, microcontroller system, software system and calibration model. The signal acquisition system was composed of three LED of different wavelength and a light-to-frequency converter. Three LED of different wavelength were lighted by turns for avoiding interaction. Light-to-frequency converter was used as the receiving tube, thus simplifying the circuit structure greatly. This paper described the instrument’s hardware design, software design, modeling of Forsythia leaf water content and forecasting. Predicted results were consistent with the true values of water, and the correlation coefficient between them was about 0.820. Advantages of this instrument were small, simple structure, low power consumption and so on. The experimental results showed that the instrument could detect plant water content rapidly on fieldwork.