Remote Sensing for Natural Resources >
A method for plastic greenhouse extraction integrating Sentinel-2 spectral indices and an improved one-class random forest
Received date: 2024-05-09
Revised date: 2024-08-01
Online published: 2026-06-03
Plastic greenhouses have gained extensive application in modern agriculture. This, however, gives rise to ecological issues. Remote sensing data enable effective extraction and identification of plastic greenhouses on a large scale. Existing studies largely focus on plastic greenhouse extraction using either classification or spectral indices methods. However, there exists a lack of the combination and comparative analysis of both methods. This study proposed a method for plastic greenhouse extraction that integrates multiple Sentinel-2 spectral indices and a one-class classification method (improved one-class random forest). Furthermore, this study extracted information on plastic greenhouses using an improved one-class random forest method, as well as six spectral indices of plastic greenhouses as classification features. The extraction results were then compared with those of the proposed method to demonstrate the effectiveness of the latter. The results indicate that the proposed method yielded an overall accuracy of above 97% across four seasons, with kappa coefficients exceeding 0.82 and F1 scores of over 0.84. These metrics all were better than those yielded using the six spectral indices. Furthermore, the proposed method exhibited differences in the overall accuracy, kappa coefficient, and F1 score across four seasons of less than 1%, under 0.1, and below 0.1 respectively. This suggests the high seasonal stability of the method, outperforming the extraction results obtained by using spectral indices alone. This study provides a method for accurately monitoring the spatial distribution of plastic greenhouses.
Key words: plastic greenhouse; spectral index; one-class random forest
XIAO Mingzhu , LI PeiJjun . A method for plastic greenhouse extraction integrating Sentinel-2 spectral indices and an improved one-class random forest[J]. Remote Sensing for Natural Resources, 2025 , 37(4) : 40 -47 . DOI: 10.6046/zrzyyg.2024159
表1 Sentinel-2波段设置及分辨率Tab.1 Sentinel-2 band settings and resolution |
| 波段 | 描述 | S-2A 中心 波长/nm | 空间分辨率/m |
|---|---|---|---|
| Band 1 | 海岸气溶胶 | 442.7 | 60 |
| Band 2 | 蓝光 | 492.4 | 10 |
| Band 3 | 绿光 | 559.8 | 10 |
| Band 4 | 红光 | 664.6 | 10 |
| Band 5 | 红边1 | 704.1 | 20 |
| Band 6 | 红边2 | 740.5 | 20 |
| Band 7 | 红边3 | 782.8 | 20 |
| Band 8 | 近红外 | 832.8 | 10 |
| Band 8A | 窄近红外 | 864.7 | 20 |
| Band 9 | 水蒸气 | 945.1 | 60 |
| Band 10 | 卷云 | 1 373.5 | 60 |
| Band 11 | 短波红外1 | 1 613.7 | 20 |
| Band 12 | 短波红外2 | 2 202.4 | 20 |
表2 各季节图像塑料大棚及非塑料大棚样本数量Tab.2 The numbers of plastic greenhouse and non-plastic greenhouse samples in images of different seasons |
| 图像 | 塑料大棚样本 数/个 | 非塑料大棚样本 数/个 |
|---|---|---|
| 春季图像 | 36 063 | 236 010 |
| 夏季图像 | 35 401 | 233 065 |
| 秋季图像 | 36 106 | 231 574 |
| 冬季图像 | 36 027 | 232 506 |
表3 本研究使用的塑料大棚指数公式Tab.3 The plastic greenhouse indices and the equations used in this study |
| 指数名 | 公式 | 变量含义 |
|---|---|---|
| VI | , , , , , , 分别为海岸气溶胶、蓝光、绿光、红光、近红外、短波红外1、短波红外2波段的光谱反射率; 与 分别为上述一组波段中最短和最长的波长; 为波长i对应波段的反射率 | |
| PMLI | ||
| MDI | ||
| GDI | ||
| PGI | ||
| APGI |
图4 研究区春季不同指数的图像Fig.4 Spectral index images from Sentinel-2 image of spring season in the study area |
表4 利用春季图像的指数与本文方法的塑料大棚提取精度Tab.4 Accuracies of plastic greenhouse extraction from Sentinel-2 image of spring season using different spectral indices and the proposed method |
| 方法 | OA/% | Kappa系数 | F1分数 | 塑料大棚 | 非塑料大棚 | ||
|---|---|---|---|---|---|---|---|
| UA/% | PA/% | UA/% | PA/% | ||||
| VI | 24.41 | -0.069 3 | 0.101 5 | 5.68 | 47.80 | 81.19 | 22.12 |
| PMLI | 70.19 | 0.108 7 | 0.231 3 | 15.03 | 50.16 | 93.65 | 72.16 |
| MDI | 86.54 | 0.418 3 | 0.486 8 | 36.94 | 71.38 | 96.91 | 88.04 |
| GDI | 81.42 | 0.316 7 | 0.403 8 | 28.31 | 70.38 | 96.60 | 82.51 |
| PGI | 93.57 | 0.555 0 | 0.589 2 | 68.81 | 51.52 | 95.36 | 97.70 |
| APGI | 95.22 | 0.720 6 | 0.746 9 | 70.96 | 78.83 | 97.90 | 96.83 |
| 本文方法 | 97.92 | 0.866 8 | 0.878 1 | 92.25 | 83.78 | 98.42 | 99.31 |
表5 利用夏、秋、冬季图像的指数与本文方法的塑料大棚提取精度Tab.5 Accuracies of plastic greenhouse extraction from Sentinel-2 image of summer, autumn and winter seasons using different indices and the proposed method |
| 方法 | OA/% | Kappa系数 | F1分数 | ||||||
|---|---|---|---|---|---|---|---|---|---|
| 夏季 | 秋季 | 冬季 | 夏季 | 秋季 | 冬季 | 夏季 | 秋季 | 冬季 | |
| VI | 54.16 | 45.47 | 11.57 | 0.013 3 | -0.021 6 | -0.197 5 | 0.155 6 | 0.147 0 | 0.089 7 |
| PMLI | 70.28 | 76.91 | 78.93 | 0.190 9 | 0.274 6 | 0.345 8 | 0.301 6 | 0.378 9 | 0.463 8 |
| MDI | 91.23 | 89.96 | 91.05 | 0.453 1 | 0.559 9 | 0.686 4 | 0.501 1 | 0.613 4 | 0.738 2 |
| GDI | 85.00 | 83.28 | 85.31 | 0.338 6 | 0.303 2 | 0.520 8 | 0.414 0 | 0.390 7 | 0.604 5 |
| PGI | 95.42 | 93.35 | 94.83 | 0.680 5 | 0.582 1 | 0.785 7 | 0.705 0 | 0.618 0 | 0.815 8 |
| APGI | 94.76 | 94.41 | 95.86 | 0.680 1 | 0.683 5 | 0.831 6 | 0.709 1 | 0.714 4 | 0.855 8 |
| 本文方法 | 97.22 | 97.64 | 97.56 | 0.829 9 | 0.858 8 | 0.898 8 | 0.845 1 | 0.871 7 | 0.912 9 |
| [1] |
|
| [2] |
|
| [3] |
郑磊, 何直蒙, 丁海勇. 基于ENVINet5的高分辨率遥感影像稀疏塑料大棚提取研究[J]. 遥感技术与应用, 2021, 36(4):908-915.
|
| [4] |
|
| [5] |
|
| [6] |
|
| [7] |
|
| [8] |
|
| [9] |
|
| [10] |
|
| [11] |
|
| [12] |
|
| [13] |
|
| [14] |
|
| [15] |
|
| [16] |
|
| [17] |
|
| [18] |
史忠奎, 李培军, 罗伦, 等. 基于形态学属性剖面和单类随机森林分类的道路路域新增建筑物提取方法[J]. 北京大学学报(自然科学版), 2018, 54(1):105-114.
|
| [19] |
|
| [20] |
|
| [21] |
|
| [22] |
|
| [23] |
|
/
| 〈 |
|
〉 |