Traditional water quality detection methods are timeconsuming and laborious， hyperspectralbased water quality monitoring enables rapid， intuitive， and insitu monitoring of water quality. In this paper， taking the Sanwan section of the ancient canal in Yangzhou as the research area， based on the combination of UAV hyperspectral and water quality navigation monitoring vessel， the waterground collaborative mode and partial least squares algorithm were used to quantitatively invert the water quality parameters such as TP， NH₃-N， TN， and I_Mn. The results show that the R² of the partial least squares inversion model is between 91%～97%. The order of fitting effect was NH₃-N>I_Mn>TN>TP. The absolute value of the inversion error of each index is between 0.2%～4%. The current method has a good inversion effect and can quickly and accurately obtain the distribution of urban river water quality， which is of great significance to urban water environment monitoring.