基于高光谱数据的密云水库水生态空间地物精细分类
doi: 10.11728/cjss2024.01.2023-0035 cstr: 32142.14.cjss2024.01.2023-0035
-
陈珠琳：女, 1994年7月出生于山东省德州市, 现为中国林业科学研究院资源信息研究所助理研究员, 主要研究方向为资源环境遥感. E-mail: chenzl@ifrit.ac.cn -
李添雨：男, 1991年8月出生于北京市. 现为北京市水科学技术研究院流域生态环境研究所工程师, 主要研究方向为水生态监测评价、水生态保护修复技术等. E-mail: lty0554@sina.com -
张耀方：女, 1991年2月出生于四川省宜宾市. 现为北京市水科学技术研究院高级工程师, 主要研究方向为水生态系统评价、水生态保护修复技术研究、流域面源污染研究等. E-mail: ucaszyf@163.com
作者简介:
通讯作者:
Land Cover Classification from Hyperspectral Data in the Water Ecological Space of Miyun Reservoir
-
摘要: 重点水域水生态空间地物类型分布状况是其健康评估以及生态规划的重要基础. 基于高分五号(GF-5)卫星高光谱数据, 采用混合式特征选择算法开展北京市密云水库水生态空间地物精细分类研究. 采用随机森林算法获取波段重要性排序, 经过特征降维将总体分类精度最高的模型对应的特征集作为初始特征子集. 利用后向序列选择算法搜索地物精细分类的最佳特征子集, 进而开展密云水库水生态空间的地物精细分类. 结果表明, 高光谱数据可以实现较高精度的地物分类(总体分类精度为93.61%, Kappa系数为91.71%), 相比于哨兵二号(S-2)卫星多光谱数据, 在精细树种分类方面具有明显的优势.Abstract: With the acceleration of China’s urbanization process, the problem of the structure and function of water ecological space has become increasingly severe. Monitoring the detailed distribution of land cover types in the key water ecological space is critical for their health assessment and future ecological planning. This study investigated a hybrid feature selection algorithm and GF-5 hyperspectral data (with a spatial resolution of 30 m) to generate a fine land cover classification method for the water ecological space of Miyun Reservoir in Beijing. Firstly, the feature importance ranking was determined using the Random Forest (RF) algorithm and several feature subsets were generated with feature amount gradually carried out in a step size of 10. Then, the classification model was generated based on each subset using the RF algorithm. The feature subset that achieved the highest overall classification accuracy was determined as the initial feature subset. Next, the backward sequential selection algorithm was used to this initial subset to search for the best feature subset. Finally, the classification model of the water ecological space of Miyun Reservoir was generated based on the best feature subset and RF algorithm. To validate the advance of GF-5 hyperspectral data, this study also developed a classification model using Sentinel-2 multispectral data (with a spatial resolution of 10 m) for comparison. The results indicated that hyperspectral data achieved high classification accuracy (overall classification accuracy of 93.61%, and Kappa coefficient of 91.71%), especially in the accurate recognition of tree species (The producer's accuracy and user's accuracy of the chestnut forest are 81.25% and 73.03%, respectively). The reflectance of shortwave infrared bands of GF-5 data has increased the differentiation between chestnut forests and other tree species. By contrast, Sentinel-2 data-based model achieved lower classification accuracy with an overall accuracy of 85.91% and a Kappa coefficient of 82.00%. This result indicated that although Sentinel-2 data has higher spatial resolution than GF-5 data, it still has difficulty identifying chestnut forests due to a lack of fine band information. The classification algorithm proposed in this study can provide accurate basic data for supporting the rational planning and management of water ecological spaces.
-
图 1 研究区地理位置以及GF-5卫星伪彩色合成影像(红为波段107, 绿为波段65, 蓝为波段41)
Figure 1. Study area location and false color image of GF-5 satellite hyperspectral data (Red: band 107. Green: band 65. Blue: band 41)
图 3 基于混合特征选择算法的分类模型构建流程
Figure 3. Classification model construction process based on hybrid feature selection algorithm
图 4 典型地物光谱曲线(a)及其NDVI分布箱形图(b)
Figure 4. Spectral curve of typical surface features (a) and NDVI distribution boxplot (b)
图 5 RF算法排序下的总体分类精度(OA)随变量个数递减的变化趋势
Figure 5. Trend of Overall Accuracy (OA) with decreasing number of variables under RF algorithm ranking
图 7 密云水库水生态空间典型地物类型分类结果
Figure 7. Classification results of typical surface features in Miyun reservoir
表 1 训练集和验证集中的样本数量
Table 1. Number of samples in training and validation set
数据集 类别 水体 不透水面及裸土 板栗林 农田和草地 其他林地 训练集(GF-5) 400 600 165 400 610 测试集(GF-5) 200 296 80 195 308 训练集(S-2) 1000 1000 600 1000 1000 测试集(S-2) 500 500 300 500 500 
下载: 导出CSV
表 2 基于GF-5和S-2数据的地物类型间J-M距离
Table 2. J-M distance between different surface features based on GF-5 data
类别 板栗林(GF-5/S-2) 农田和草地(GF-5/S-2) 不透水面(GF-5/S-2) 其他林地(GF-5/S-2) 水体(GF-5/S-2) 板栗林(GF-5/S-2) 0/0 1.7260/1.7861 1.9405/1.9501 1.5003/0.8491 1.9667/1.9999 农田及草地(GF-5/ S-2) 1.7260/1.7861 0/0 1.9022/1.9855 1.7470/1.7667 1.9662/1.9999 不透水面及裸土
(GF-5/ S-2)1.9405/1.9501 1.9022/1.9855 0/0 1.9484/1.9807 1.9644/1.9999 其他林地(GF-5/ S-2) 1.5003/0.8491 1.7470/1.7667 1.9484/1.9807 0/0 1.9679/1.9999 水体(GF-5/ S-2) 1.9667/1.9999 1.9662/1.9999 1.9644/1.9999 1.9679/1.9999 0/0
下载: 导出CSV
表 3 GF-5验证集分类混淆矩阵
Table 3. Confusion matrix of GF-5 verification set
类别 水体 不透水面和裸土 板栗林 农田和草地 其他林地 PA/(%) 水体 200 0 0 0 0 100 不透水面及裸土 0 290 0 6 0 97.97 板栗林 0 0 65 2 13 81.25 农田和草地 0 10 3 175 7 89.74 其他林地 0 0 21 7 280 90.91 UA/(%) 100 96.67 73.03 92.11 93.33 － 注 UA代表用户精度, PA代表生产者精度.
下载: 导出CSV
表 4 S-2验证集分类混淆矩阵
Table 4. Confusion matrix of S-2 verification set
类别 水体 不透水面和裸土 板栗林 农田和草地 其他林地 PA/(%) 水体 500 0 0 0 0 100.00 不透水面及裸土 0 497 0 3 0 98.61 板栗林 0 0 0 0 300 0.00 农田和草地 0 7 0 484 9 98.37 其他林地 0 0 0 5 495 61.57 UA/(%) 100 99.40 0 96.80 99.00 －
下载: 导出CSV
-
[1] 周广金, 童亚莉, 王凌青, 等. 国土空间规划中水生态空间及保护线的多维识别技术与应用[J]. 自然资源学报, 2022, 37(12): 3102-3117 doi: 10.31497/zrzyxb.20221206ZHOU Guangjin, TONG Yali, WANG Lingqing, et al. Multi-dimensional identification technology and application of water ecological space and protection line in the territorial spatial planning[J]. Journal of Natural Resources, 2022, 37(12): 3102-3117 doi: 10.31497/zrzyxb.20221206 [2] 周启刚, 李剑, 孟浩斌, 等. 基于遥感解释的重庆市重要生态空间质量评价[J]. 水土保持研究, 2021, 28(6): 292-300ZHOU Qigang, LI Jian, MENG Haobin, et al. Evaluation on quality of important ecological space in Chongqing city based on remote sensing interpretation[J]. Research of Soil and Water Conservation, 2021, 28(6): 292-300 [3] 朱琦, 郭华东, 张露, 等. 基于多时相Landsat8影像的海南岛热带天然林类型遥感分类[J]. 自然资源遥感, 2022, 34(2): 215-223ZHU Qi, GUO Huadong, ZHANG Lu, et al. Classification of tropical natural forests in Hainan Island based on multi-temporal Landsat8 remote sensing images[J]. Remote Sensing for Natural Resources, 2022, 34(2): 215-223 [4] 王林江, 吴炳方, 张淼, 等. 关键生育期冬小麦和油菜遥感分类方法[J]. 地球信息科学学报, 2019, 21(7): 1121-1131 doi: 10.12082/dqxxkx.2019.180421WANG Linjiang, WU Bingfang, ZHANG Miao, et al. Winter wheat and rapeseed classification during key growth period by integrating multi-source remote sensing data[J]. Journal of Geo-Information Science, 2019, 21(7): 1121-1131 doi: 10.12082/dqxxkx.2019.180421 [5] 苏红军, 刘浩. 一种利用空间和光谱信息的高光谱遥感多分类器动态集成算法[J]. 国土资源遥感, 2017, 29(2): 15-21SU Hongjun, LIU Hao. A novel dynamic classifier selection algorithm using spatial-spectral information for hyperspectral classification[J]. Remote Sensing for Land & Resources, 2017, 29(2): 15-21 [6] 韩文军, 张苏, 焦全军, 等. 基于多时相CHRIS高光谱卫星数据的优势树种分类研究[J]. 林业调查规划, 2019, 44(2): 1-6 doi: 10.3969/j.issn.1671-3168.2019.02.001HAN Wenjun, ZHANG Su, JIAO Quanjun, et al. Dominant tree species mapping based on multi-temporal CHRIS hyperspectral satellite data[J]. Forest Inventory and Planning, 2019, 44(2): 1-6 doi: 10.3969/j.issn.1671-3168.2019.02.001 [7] 国强, 彭龙. 基于三维卷积神经网络与超像素分割的高光谱分类[J]. 光学学报, 2021, 41(22): 2210001 doi: 10.3788/AOS202141.2210001GUO Qiang, PENG Long. Hyperspectral classification based on 3D convolutional neural network and super pixel segmentation[J]. Acta Optica Sinica, 2021, 41(22): 2210001 doi: 10.3788/AOS202141.2210001 [8] BHOSLE K, MUSANDE V. Evaluation of deep learning CNN model for land use land cover classification and crop identification using hyperspectral remote sensing images[J]. Journal of the Indian Society of Remote Sensing, 2019, 47(11): 1949-1958 doi: 10.1007/s12524-019-01041-2 [9] DING X H, ZHANG S Q, LI H P, et al. A restrictive polymorphic ant colony algorithm for the optimal band selection of hyperspectral remote sensing images[J]. International Journal of Remote Sensing, 2020, 41(3): 1093-1117 doi: 10.1080/01431161.2019.1655810 [10] LI J D, CHENG K W, WANG S H, et al. Feature selection: A data perspective[J]. ACM Computing Surveys, 2017, 50(6): 94 [11] WEI G F, ZHAO J, FENG Y L, et al. A novel hybrid feature selection method based on dynamic feature importance[J]. Applied Soft Computing, 2020, 93: 106337 doi: 10.1016/j.asoc.2020.106337 [12] CHEN Z L, JIA K, XIAO C C, et al. Leaf area index estimation algorithm for GF-5 hyperspectral data based on different feature selection and machine learning methods[J]. Remote Sensing, 2020, 12(13): 2110 doi: 10.3390/rs12132110 [13] 张立福, 赵晓阳, 孙雪剑, 等. 高分五号高光谱数据融合方法比较[J]. 遥感学报, 2022, 26(4): 632-645 doi: 10.11834/jrs.20229318ZHANG Lifu, ZHAO Xiaoyang, SUN Xuejian, et al. Comparison of fusion methods on GF-5 hyperspectral data[J]. National Remote Sensing Bulletin, 2022, 26(4): 632-645 doi: 10.11834/jrs.20229318 [14] 韩冰冰, 陈圣波, 曾庆鸿, 等. 基于J-M距离的多时相Sentinel-1农作物分类[J]. 科学技术与工程, 2020, 20(17): 6977-6982 doi: 10.3969/j.issn.1671-1815.2020.17.040HAN Bingbing, CHEN Shengbo, ZENG Qinghong, et al. Time-series classification of Sentinel-1 data based on J-M distance[J]. Science Technology and Engineering, 2020, 20(17): 6977-6982 doi: 10.3969/j.issn.1671-1815.2020.17.040 [15] 刘咏梅, 盖星华, 董幸枝, 等. 退化高寒草甸主要物种花与叶片的光谱识别方法[J]. 西北大学学报(自然科学版), 2022, 52(2): 159-168LIU Yongmei, GE Xinghua, DONG Xingzhi, et al. Spectral discrimination of flowers and leaves of major species in the degraded alpine meadow[J]. Journal of Northwest University (Natural Science Edition), 2022, 52(2): 159-168 [16] 牛全福, 傅键恺, 陆铭, 等. 基于随机森林的GF-6 WFV和Landsat8 OLI遥感影像分类比较[J]. 地理空间信息, 2022, 20(8): 49-54 doi: 10.3969/j.issn.1672-4623.2022.08.012NIU Quanfu, FU Jiankai, LU Ming, et al. Comparison of GF-6 WFV and Landsat8 OLI remote sensing image classification based on random forest[J]. Geospatial Information, 2022, 20(8): 49-54 doi: 10.3969/j.issn.1672-4623.2022.08.012 [17] BIAU G, SCORNET E. A random forest guided tour[J]. Test, 2016, 25(2): 197-227 doi: 10.1007/s11749-016-0481-7 [18] 曹泽涛, 方子东, 姚瑾, 等. 基于随机森林的黄土地貌分类研究[J]. 地球信息科学学报, 2020, 22(3): 452-463 doi: 10.12082/dqxxkx.2020.190247CAO Zetao, FANG Zidong, YAO Jin, et al. Loess landform classification based on random forest[J]. Journal of Geo-information Science, 2020, 22(3): 452-463 doi: 10.12082/dqxxkx.2020.190247 [19] 杨红艳, 杜健民, 阮培英, 等. 基于无人机遥感与随机森林的荒漠草原植被分类方法[J]. 农业机械学报, 2021, 52(6): 186-194 doi: 10.6041/j.issn.1000-1298.2021.06.019YANG Hongyan, DU Jianmin, RUAN Peiying, et al. Vegetation classification of desert steppe based on unmanned aerial vehicle remote sensing and random forest[J]. Transactions of the Chinese Society for Agricultural Machinery, 2021, 52(6): 186-194 doi: 10.6041/j.issn.1000-1298.2021.06.019 [20] AHMED K A, ALJAHDALI S, HUSSAIN S N. Comparative prediction performance with support vector machine and random forest classification techniques[J]. International Journal of Computer Applications, 2013, 69(11): 12-16 doi: 10.5120/11885-7922 [21] RUAN F Q, QI J, YAN C H, et al. Quantitative detection of harmful elements in alloy steel by LIBS technique and sequential backward selection-random forest (SBS-RF)[J]. Journal of Analytical Atomic Spectrometry, 2017, 32(11): 2194-2199 doi: 10.1039/C7JA00231A [22] GUNASUNDARI S, JANAKIRAMAN S. A hybrid PSO-SFS-SBS algorithm in feature selection for liver cancer data[M]//Power Electronics and Renewable Energy Systems. New Delhi: Springer, 2015: 1369-1376 [23] 陈珠琳, 贾坤, 李强子, 等. 基于混合式特征选择的高分五号影像农田识别[J]. 遥感学报, 2022, 26(7): 1383-1394 doi: 10.11834/jrs.20220458CHEN Zhulin, JIA Kun, LI Qiangzi, et al. Hybrid feature selection for cropland identification using GF-5 satellite image[J]. National Remote Sensing Bulletin, 2022, 26(7): 1383-1394 doi: 10.11834/jrs.20220458 -
-
下载:
图(7) / 表(4)
计量
- 文章访问数: 466
- HTML全文浏览量: 101
- PDF下载量: 70
-
被引次数:
0(来源:Crossref)
0(来源:其他)
