To obtain real-time and accurate ammonia nitrogen content in aquaculture seawater, taking ammonia nitrogen as the research object, the colorimetric reaction of indophenol blue method was carried out on the basis of colorimetric spectroscopy combined with microfluidic technology to realize the quantitative detection of ammonia nitrogen in solution. After processing the data, a corresponding model was established and the effects of different spectral preprocessing and different sample set partitioning algorithms on the establishment of the detection model were compared, in which after the multiplication scatter correction, the partial least squares (PLS) regression model established by the wavelet smoothing preprocessing method combined with the sorting method to divide the sample set had the best effect. The root mean standard error of calibration (RMSEC) and the root mean standard error of prediction (RMSEP) were respectively 0.0566mg/L and 0.0677mg/L, the relative analytical error (RPD) was 6.8932;the linear range and detection limit of the method measured under optimized conditions were 0.005~1.350mg/L and 0.0036mg/L, respectively. The standard recovery experiment was carried out on seawater, tap water and acquaculture water bodies. The average recovery rate was between 94% and 109%, and the relative standard deviation was between 2.3% and 5.8%. The results showed that the experimental modeling effect was good, the operation was simple and convenient, and the experiment was fast, reliable, and pollution-free. It showed that the method of detecting ammonia nitrogen using colorimetric spectroscopy combined with microfluidic technology was feasible.