| 基于遥感影像与逻辑回归模型的延河流域沟壑分布概率预测 |
| |
| 引用本文： | 范天程,贾云飞,李云飞,赵建林.基于遥感影像与逻辑回归模型的延河流域沟壑分布概率预测[J].水土保持研究,2022,29(4):316-321. |
| |
| 作者姓名： | 范天程 贾云飞 李云飞 赵建林 |
| |
| 作者单位： | (长安大学 地质工程与测绘学院, 西安 710054) |
| |
| 摘 要： | 为研究黄土高原沟壑地貌空间分布特征,开展了基于遥感影像和机器学习的沟壑地貌提取,研究了延河流域沟壑地貌空间分布及其环境控制因子。以Google Earth Pro平台为支撑,在人工提取大量沟壑地貌样本的基础上,基于Landsat8 OLI影像波段信息,采用主成分分析前3个变量、缨帽变换前3个变量、NDVI、高程、坡度和坡向10个因子,使用逻辑回归模型预测整个延河流域的沟壑概率分布。结果表明:(1)在10个变量因子中,因子Brightness的R2McF为0.158,重要性最大,因子elevation的R2McF为3.6×10-5,重要性最小;(2)最优逻辑回归模型由组合因子Brightness,PCA1,Greenness,Wetness,PCA3和slope确定,其重要性R2McF为0.206;(3)最优逻辑回归模型的沟壑概率预测精度为73.72%,ROC曲线下面积即AUC值为0.80;(4)延河流域沟壑地貌约占整个延河流域面积的52.05%。研究表明,延河流域沟壑分布呈现从西北方向到东南方向逐渐集中的特点。
|
| 关 键 词： | 遥感影像 沟壑分布 逻辑回归模型 多因子 延河流域 黄土高原 |
Prediction of Gully Distribution Probability in Yanhe Basin Based on Remote Sensing Image and Logistic Regression Model |
| |
| FAN Tiancheng,JIA Yunfei,LI Yunfei,ZHAO Jianlin.Prediction of Gully Distribution Probability in Yanhe Basin Based on Remote Sensing Image and Logistic Regression Model[J].Research of Soil and Water Conservation,2022,29(4):316-321. |
| |
| Authors: | FAN Tiancheng JIA Yunfei LI Yunfei ZHAO Jianlin |
| |
| Institution: | (School of Geological Engineering and Surveying and Mapping, Chang'an University, Xi'an 710054, China) |
| |
| Abstract: | In order to study the spatial distribution characteristics of gully landform on the Loess Plateau, remote sensing images and machine learning were used to extract gully landform, and the spatial distribution and environmental control factors of gully landform in the Yanhe Basin were studied. We estimated the probability of gully distribution based on a large number of gully samples extracted manually from Google Earth Pro platform and the logical regression model. Based on the Landsat8-OLI image, we established 10 environmental and spectrum factors, including the first three variables of principal component analysis, the first three variables of Landsat8-OLI transform, NDVI, elevation, slope as well as aspect, to calibrate the logistic regression method. Then, the model was used to predict the probability distribution of gullies in the entire Yanhe Basin. The results show that:(1)among the 10 variables, the most important factor is brightness with the R2McF of 0.158, while elevation has the lowest contribution with the R2McF of 3.6×10-5;(2)the optimal logistic regression model is determined by the combined factors of Brightness, PCA1, Greenness, Wetness, PCA3 and slope, and its overall R2McF is 0.206;(3)the accuracy of the optimal logistic regression model is 73.72% and the AUC value of the ROC curve is 0.80, which indicates that the model has relatively higher accuracy to estimate the distribution of gully on the Loess Plateau;(4)the gully landform accounts for 52.05% of the total areas of Yanhe Basin. The study shows that the distribution of gullies in Yanhe Basin gradually concentrates from northwest to southeast. |
| |
| Keywords: | remote sensing image gully distribution logistic regression model multiple-factor Yanhe Basin Loess Plateau |
|
| 点击此处可从《水土保持研究》浏览原始摘要信息 |
| 点击此处可从《水土保持研究》下载免费的PDF全文 |
|
