基于竞争适应重加权采样算法耦合机器学习的土壤含水量估算 下载: 965次
Estimation of Soil Moisture Content Based on Competitive Adaptive Reweighted Sampling Algorithm Coupled with Machine Learning
葛翔宇 1,2,3,*丁建丽 1,2,3,*王敬哲 1,2,3王飞 1,2,3蔡亮红 1,2,3孙慧兰 4
1 新疆大学资源与环境科学学院, 新疆 乌鲁木齐 830046
2 新疆大学绿洲生态教育部重点实验室, 新疆 乌鲁木齐 830046
3 新疆大学智慧城市与环境建模自治区普通高校重点实验室, 新疆 乌鲁木齐 830046
4 新疆师范大学地理科学与旅游学院, 新疆 乌鲁木齐 830054
图 & 表
图 1. 全样本MCCV土壤预测残差的均值-标准差分布
Fig. 1. Mean-standard deviation distribution of soil-residual prediction for full-sample MCCV
下载图片 查看原文
图 2. 实验及模型计算过程流程图
Fig. 2. Flow chart of calculation process for experience and model
下载图片 查看原文
图 3. 不同SMC土壤的光谱反射率
Fig. 3. Spectral reflectance of soils with different SMCs
下载图片 查看原文
图 4. CARS算法筛选变量的过程。(a)波长变量个数的变化;(b) RMSECV的变化;(c) RMSECV最小时变量回归系数的趋势
Fig. 4. Variable filtering process using CARS. (a) Variation in wavelength variable number; (b) variation in RMSECV; (c) trend of variable regression coefficient when RMSECV is minimum
下载图片 查看原文
图 5. 土壤样本反射率均值及最优光谱波段
Fig. 5. Mean reflectance of soil samples and optimal spectral bands
下载图片 查看原文
图 6. ELM模型SMC的预测值与实测值
Fig. 6. Predicted and measured SMCs using ELM model
下载图片 查看原文
表 1土壤样品的SMC统计特征
Table1. Statistical characteristics of SMC of soil samples
Sample type | Number | Maximum | Minimum | Mean | Standard deviation | Coefficient of variation |
---|
Whole set | 77 | 0.252 | 0.021 | 0.1421 | 0.049 | 0.3458 | Calibration set | 62 | 0.252 | 0.021 | 0.1406 | 0.051 | 0.3659 | Validation set | 15 | 0.216 | 0.067 | 0.1483 | 0.039 | 0.2637 |
|
查看原文
表 2SMC预测结果
Table2. Estimated SMC
Model | Variable number | Calibration set | Prediction set |
---|
RMSE | R2 | | RMSE | R2 | RPD | RPIQ |
---|
PLSR | 20 | 0.484 | 0.478 | 0.622 | 0.617 | 0.522 | 0.18401 | BPNN | 20 | 0.027 | 0.706 | 0.024 | 0.799 | 2.016 | 1.90200 | RFR | 20 | 0.024 | 0.872 | 0.021 | 0.898 | 1.647 | 2.18900 | ELM | 20 | 0.016 | 0.879 | 0.015 | 0.918 | 3.123 | 3.32500 |
|
查看原文
表 3不同建模预测比的SMC预测结果
Table3. Predicted SMC based on different ratios of calibration to prediction
Model | Ratio of calculation to prediction | Variable number | Calibration set | Prediction set |
---|
RMSE | R2 | | RMSE | R2 | RPD | RPIQ |
---|
BPNN | 62∶15 | 20 | 0.027 | 0.706 | 0.024 | 0.799 | 2.016 | 1.902 | 57∶20 | 20 | 0.020 | 0.842 | 0.023 | 0.800 | 1.826 | 1.499 | 52∶25 | 20 | 0.023 | 0.765 | 0.024 | 0.800 | 1.947 | 2.010 | RFR | 62∶15 | 20 | 0.024 | 0.872 | 0.021 | 0.898 | 1.647 | 2.189 | 57∶20 | 20 | 0.024 | 0.863 | 0.013 | 0.897 | 2.217 | 3.073 | 52∶25 | 20 | 0.025 | 0.856 | 0.014 | 0.889 | 2.202 | 3.041 | ELM | 62∶15 | 20 | 0.016 | 0.879 | 0.015 | 0.918 | 3.123 | 3.325 | 57∶20 | 20 | 0.019 | 0.869 | 0.014 | 0.919 | 3.102 | 3.241 | 52∶25 | 20 | 0.015 | 0.877 | 0.016 | 0.918 | 2.569 | 2.958 |
|
查看原文
葛翔宇, 丁建丽, 王敬哲, 王飞, 蔡亮红, 孙慧兰. 基于竞争适应重加权采样算法耦合机器学习的土壤含水量估算[J]. 光学学报, 2018, 38(10): 1030001. Xiangyu Ge, Jianli Ding, Jingzhe Wang, Fei Wang, Lianghong Cai, Huilan Sun. Estimation of Soil Moisture Content Based on Competitive Adaptive Reweighted Sampling Algorithm Coupled with Machine Learning[J]. Acta Optica Sinica, 2018, 38(10): 1030001.