基于多方位DEM地形晕渲的黄土地貌正负地形提取
作者简介: 陈永刚(1980-),男,内蒙兴和人,博士研究生,从事DEM信息提取研究及GIS在林业中应用开发工作。E-mail:cyg_gis@163.com
收稿日期: 2011-02-20
要求修回日期: 2011-04-11
网络出版日期: 2012-01-20
基金资助
国家自然科学基金项目(40930531、40801148)、浙江省教育厅项目(Y201017891)资助
The Positive and Negative Terrain of Loess Plateau Extraction Based on the Multi-azimuth DEM Shaded Relief
Received date: 2011-02-20
Request revised date: 2011-04-11
Online published: 2012-01-20
Copyright
以陕北绥德县韭园沟流域5 m分辨率DEM数据为基础,在数字地形分析、多元统计和数据挖掘方法的基础上,提出利用多方位DEM地形晕渲、坡度等多元指标,以主成分分析消除多重共线性和约减维数,并以Logistic回归模型提取黄土高原正、负地形的方法。研究结果表明:模型提取精度为82.1%,Kappa统计量为0.629;模型在6个不同流域测试样本上正、负地形的平均精度分别为77.6%,84.9%,加权平均精度为81.3%,模型具有较好的一致性和泛化能力,正、负地形提取效果良好。
关键词: 正负地形; 黄土高原; 主成分分析; Logistic回归
陈永刚 , 汤国安 , 周毅 , 李发源 , 宴实江 , 张磊 . 基于多方位DEM地形晕渲的黄土地貌正负地形提取[J]. 地理科学, 2012 , 32(1) : 105 -109 . DOI: 10.13249/j.cnki.sgs.2012.01.105
Based on the DEM data of Jiuyuangou wateshed in Suide County, Shaanxi of China, with a spatial resolution of 5 m, employing the digital terrain analysis, multivariate statistics and data mining methods, multiple indexes of multi-azimuth DEM shaded relief and slope are established. The solutions of extraction positive and negative terrain of loess plateau by Principle Components Analysis, and Logistic regression model is proposed. The result indicates that: the extracting model has better consistency and accuracy, of which the accuracy is 82.1% and Kappa statistics is 0.6298. Tested on 6 samples of different valley by this model, the mean accuracy of positive and negative terrain are at 77.6% and 84.9%, and weighed mean accuracy is 81.3%. It is suitable to extract positive and negative terrain of loess plateau by PCA and Logistic regression model.
Fig.1 Sketch of tudy area图1 研究区 |
Fig. 2 Principal component screeplot (a) and loading (b)图2 主成分碎石(a)、载荷(b) |
Table 1 Eigenvalue and contribution rate表1 特征值与贡献率 |
项目 | 1 | 2 | 3 | 4 | 5 | … | 17 | 18 | 19 |
---|---|---|---|---|---|---|---|---|---|
特征值 | 7.6459 | 6.5736 | 3.8480 | 0.8922 | 0.0395 | … | 0.0000 | 0.0000 | 0.0000 |
贡献率 | 0.4020 | 0.3460 | 0.2030 | 0.0470 | 0.0020 | … | 0.0000 | 0.0000 | 0.0000 |
累计贡献率 | 0.4020 | 0.7480 | 0.9510 | 0.9980 | 1.0000 | … | 1.0000 | 1.0000 | 1.0000 |
Table 2 Logistic regression coefficients and significance test表2 Logistic回归系数及显著性检验 |
变量名 | 系数 | 标准差 | Z值 | 概率P |
---|---|---|---|---|
常数项 | 0.5471 | 0.0136 | 40.240 | 0.000 |
Y1 | 0.5950 | 0.0063 | 94.300 | 0.000 |
Y2 | 0.0802 | 0.0051 | 15.680 | 0.000 |
Y3 | 0.0842 | 0.0066 | 12.680 | 0.000 |
Y4 | 1.6539 | 0.0178 | 92.920 | 0.000 |
Table 3 Model accuracy evaluation表3 模型精度评价 |
项 目 | TP | FP | Precision | Recall | F-Measure | ROC |
---|---|---|---|---|---|---|
正地形 | 0.868 | 0.244 | 0.832 | 0.868 | 0.85 | 0.897 |
负地形 | 0.756 | 0.132 | 0.805 | 0.756 | 0.78 | 0.897 |
加权平均值 | 0.821 | 0.197 | 0.821 | 0.821 | 0.82 | 0.897 |
Table 4 Accuracy of testing samples表4 测试样区精度 |
测试样区 | 正地形精度 | 负地形精度 | 加权平均精度 |
---|---|---|---|
1 | 0.788 | 0.855 | 0.820 |
2 | 0.809 | 0.809 | 0.809 |
3 | 0.703 | 0.725 | 0.713 |
4 | 0.748 | 0.951 | 0.862 |
5 | 0.819 | 0.882 | 0.848 |
6 | 0.787 | 0.872 | 0.826 |
平均值 | 0.776 | 0.849 | 0.813 |
Fig.3 Comparison of study area图3 研究样区对比 |
The authors have declared that no competing interests exist.
[1] |
|
[2] |
|
[3] |
|
[4] |
|
[5] |
|
[6] |
|
[7] |
|
[8] |
|
[9] |
|
[10] |
|
[11] |
|
[12] |
|
[13] |
|
[14] |
|
[15] |
|
[16] |
|
/
〈 | 〉 |