In order to realize the precise automatic irrigation of melon cultivation in facilities, an irrigation decision-making system based on the growing demand of melon was designed. Taking into account the three categories of melon yield, quality, and water and nitrogen use efficiency, a comprehensive evaluation system was established involving ten indicators, and the AHP method integrating the maximum membership degree was adopted to determine the comprehensive optimal daily irrigation level for melon cultivation. The K-means clustering algorithm was used to analyze the daily optimal irrigation amount and ambient temperature and humidity, and a quantitative irrigation decision model based on the clustering results of vapor pressure deficit (VPD) was established for melon in different growth stages. The results showed that the comprehensive growth of melon was optimal under the irrigation level of 120%evapotranspiration. When the number of cluster centroids was 3, the contour coefficient was the largest of 0.72. The contours and boundaries between groups were clear, and the irrigation amount was increased significantly with the increase in VPD, and the clustering results were the best. For the automatic irrigation system, RS485 sensors of temperature and humidity were used to detect the greenhouse environmental parameters in real time, and model-based irrigation control was realized by microcontroller of STM32F103ZET6, and electronic flowmeter was used to measure and feedback the amount of irrigation water, and then remote monitoring function was realized through cloud platform. The system application verification test showed that the melon produced by this system was slightly better than that of conventional agronomic management in yield, soluble solids and soluble protein, and it had significant advantages in water saving, with 15.9% reduction in irrigation amount for the whole growth period, which also greatly reduced labor costs, and realized automatic precision irrigation.