A water quality parameter inversion model was established based on machine learning and by using UAV multispectral reflectance image to detect water quality.Applying this model to inverting the water quality of Zhangjiagang River,the results indicated that random forest algorithm had high accuracy in inver-ting ammonia nitrogen on the optimal band combination selected by feature variable importance analysis based on XGBoost and random forest.The determination coefficient of the test set was 0.91,and the average absolute per-centage error was 23.57％.The inversion results could accurately reflect the water quality characteristics of the tributaries in Guangmingcun section of Zhangjiagang River in space,and visually displayed the key areas where water quality exceeding the standard.
维普
万方数据