中国南方典型湿润山区植被类型的无人机多光谱遥感机器学习分类研究

张妮娜,张珂,李运平,李曦,刘涛

Study on Machine Learning Methods for Vegetation Classification in Typical Humid Mountainous Areas of South China based on the UAV Multispectral Remote Sensing

Nina ZHANG,Ke ZHANG,Yunping LI,Xi LI,Tao LIU

表2 基于不同机器学习模型的分类精度评价

Table 2 Comparison of the classification accuracy of the four machine learning methods

类别 DT RF SVM AdaBoost UA/% PA/% UA/% PA/% UA/% PA/% UA/% PA/% 杉树 80.31 88.70 87.39 90.43 83.08 85.71 90.37 94.57 竹林 90.83 83.19 90.32 94.12 79.66 78.33 97.54 96.74 杉竹混合林 94.51 92.47 96.67 93.55 78.95 80.65 97.75 95.60 乔木混合林 96.10 97.37 97.33 96.05 95.83 92.00 97.50 98.73 灌木丛 100.00 76.47 100.00 64.71 92.31 85.71 100.00 100.00 农作物 80.26 94.74 90.48 100.00 92.31 92.31 100.00 76.92 裸地 83.33 71.43 100.00 85.71 57.14 100.00 100.00 66.67 水体 100.00 100.00 100.00 100.00 100.00 25.00 100.00 100.00 总体精度/% 89.87 92.43 83.52 95.55 Kappa系数 0.861 6 0.903 7 0.787 6 0.941 9 均方误差 0.305 1 0.405 3 0.514 5 0.376 4