結合多尺度分割和隨機森林的變質礦物提取

唐淑蘭; 孟勇; 王國強; 卜濤

doi:10.13374/j.issn2095-9389.2020.09.08.004

結合多尺度分割和隨機森林的變質礦物提取

doi: 10.13374/j.issn2095-9389.2020.09.08.004

唐淑蘭¹,
孟勇^2, ,,
王國強²,
卜濤²

1.
西安財經大學管理學院，西安 710100
2.
中國地質調查局西安地質調查中心，西安 710054

基金項目: 中國地質調查局資助項目（DD20179403，DD20190364，DD20190812）；西安財經大學科學研究扶持計劃資助項目（21FCJH008）；陜西省自然科學基礎研究計劃資助項目（2020JM-585）

詳細信息

通訊作者:
E-mail：16392800@qq.com

計量
- 文章訪問數: 739
- HTML全文瀏覽量: 454
- PDF下載量: 27
- 被引次數: 0
出版歷程
- 收稿日期: 2020-09-08
- 網絡出版日期: 2021-01-08
- 刊出日期: 2022-02-15

Extraction of metamorphic minerals by multiscale segmentation combined with random forest

1.
School of Management, Xi’an University of Finance and Economics, Xi’an 710100, China
2.
Xi’an Center of Geological Survey, China Geological Survey, Xi’an 710054, China

More Information

Corresponding author: E-mail: 16392800@qq.com

摘要

摘要: 為提高遙感影像變質礦物提取精度，提升變質帶的識別效果，以甘肅北山ASTER影像為研究區，結合了比值運算、多尺度分割、隨機森林分類法進行變質礦物提取。首先，通過礦物特征性光譜特征構造比值運算公式、進行影像增強；然后，對增強影像進行基于光譜及變差函數的多尺度分割；接著，采用隨機森林法提取目標礦物；最后，通過野外勘查、采樣、薄片鑒定進行精度評價。結果表明，黑云母、白云母、角閃石在ASTER影像上具有鑒定性特征，提取精度分別為85.4088%、84.7640%和85.7308%；其他含量較少的變質礦物提取精度可達到60%以上。多尺度分割能充分利用礦物的叢集特征；變差函數紋理能增強形態特征對礦物的區分能力；隨機森林分類法對礦物混合引起的噪聲不敏感、提取結果穩定。
- 變差函數 /
- 多尺度分割 /
- ASTER /
- 礦物提取 /
- 隨機森林
Abstract: The identification of metamorphic minerals is the basis of metamorphic rock research. Extraction of mineral information by remote sensing technology has been widely used. Digital image processing technology is also effectively applied to remote sensing image processing. Results show that the band ratio of remote sensing images can enhance mineral information, while the variogram function can describe the spatial correlation and variability of image pixels and extract more detailed texture information. The metamorphic minerals are found to present a block or strip distribution. The object-oriented remote sensing image information extraction method can avoid the “salt and pepper phenomenon” based on pixel extraction. Meanwhile, the random forest classification method has a fast calculation speed and high parameter accuracy. It is not sensitive to the noise caused by more lithologic components and its classification effect is found to be stable. To improve the extraction accuracy of metamorphic minerals from remote sensing images and further improve the recognition effect of metamorphic zones, this paper combined the ratio operation, multiscale segmentation, and random forest classification to extract metamorphic mineral information from ASTER images in Beishan area in Gansu Province. Initially, the image was enhanced by the ratio formula of the characteristic spectral structure of the target mineral. Multiscale image segmentation was then performed based on the spectrum and variogram. Finally, the accuracy was evaluated by the thin film identification results of the field exploration samples after the extraction of the target mineral by random forest. Results show that biotite, muscovite, and amphibole have identification characteristics on the ASTER image with an extraction accuracy of 85.4088%, 84.7640%, and 85.7308%, respectively. The extraction accuracy of other metamorphic minerals with less content are found to reach more than 60%. Multiscale segmentation can make full use of the clustering features of minerals and the variogram texture can enhance the ability of morphological features to distinguish the minerals. Random forest is not sensitive to noise and the extraction results are observed to be stable.
- variogram /
- multiscale segmentation /
- ASTER /
- mineral extraction /
- random forest

HTML全文

圖 1 研究區地質簡圖

Figure 1. Geological sketch of the study area

下載: 全尺寸圖片幻燈片

圖 2 標志性礦物反射率曲線

Figure 2. Reflectance curve of marker minerals

下載: 全尺寸圖片幻燈片

圖 3 技術流程

Figure 3. Technical process

下載: 全尺寸圖片幻燈片

圖 4 比值增強結果。（a）Bi；（b）Mus；（c）Am；（d）Chl；（e）Gt；（f）Act

Figure 4. Ratio enhancement results: (a) Bi; (b) Mus; (c) Am; (d) Chl; (e) Gt; (f) Act

下載: 全尺寸圖片幻燈片

圖 5 多尺度分割流程

Figure 5. Multiscale segmentation process

下載: 全尺寸圖片幻燈片

圖 6 GS值變化趨勢

Figure 6. Change trend of GS values

下載: 全尺寸圖片幻燈片

圖 7 多尺度分割結果。（a）Bi；（b）Mus；（c）Am；（d）Chl；（e）Gt；（f）Act

Figure 7. Multiscale segmentation results: (a) Bi; (b) Mus; (c) Am; (d) Chl; (e) Gt; (f) Act

下載: 全尺寸圖片幻燈片

圖 8 特征重要程度排名

Figure 8. Ranking of feature importance

下載: 全尺寸圖片幻燈片

圖 9 決策樹個數選擇

Figure 9. Number selection of decision trees

下載: 全尺寸圖片幻燈片

圖 10 礦物提取結果。（a）Bi；（b）Mus；（c）Am；（d）Chl；（e）Gt；（f）Act

Figure 10. Mineral extraction results: (a) Bi; (b) Mus; (c) Am; (d) Chl; (e) Gt; and (f) Act

下載: 全尺寸圖片幻燈片

圖 11 部分樣品顯微照片。（a）云母石英片巖；（b）石榴石白云母斜長片麻巖；（c）蛇紋石透輝石白云質大理巖；（d）鈉長石綠泥石石英千枚巖

Figure 11. Micrograph of some samples: (a) mica quartz schist; (b) garnet muscovite plagioclase gneiss; (c) serpentine diopside dolomitic marble; (d) albite chlorite quartz phyllite

Q—Quartz；Pl—Plagioclase；Kf—K-feldspar；Ser—Sericite；Am—Amphibole；Bi—Biotite；Di—Diopside；Act—Actinolite；Serp—Serpentine，Cc—Carbonate minerals；Mus—Muscovite；Gt—Garnet；Chl—Chlorite；Ep—Epidote

下載: 全尺寸圖片幻燈片

表 1 礦物的吸收譜帶與ASTER波段的對應關系

Table 1. Correspondence relation between the absorption bands of the minerals and ASTER bands

Mineral	ASTER band
Mineral	1	5	6	7	8	9	10	11	12	13
Bi							Reflex	Absorption	Reflex
Mus	Absorption	Reflex	Absorption	Reflex
Am				Reflex	Absorption	Reflex
Chl	Absorption	Reflex			Absorption	Reflex
Gt									Absorption	Reflex
Act			Reflex		Absorption	Reflex

下載: 導出CSV

表 2 各礦物比值公式

Table 2. Ratio formula of minerals

Bi	Mus	Am	Chl	Gt	Act
(b₁₂+b₁₀)/b₁₁	(b₅+b₇)/b₆	(b₆+b₉)/(b₈+b₇)	(b₁+b₉)/b₈	b₁₃/b₁₂	(b₆+ b₉)/b₈

下載: 導出CSV

表 3 礦物樣本數

Table 3. Mineral samples

Bi	Mus	Am	Chl	Gt	Act
1577	1348	1945	1403	912	832

下載: 導出CSV

表 4 不同特征數量礦物提取精度

Table 4. Extraction precision of minerals with different characteristic numbers

Features	Extraction precision/%
Features	Bi	Mus	Am	Chl	Gt	Act
1	80.0041	75.1212	72.3755	66.0052	64.6564	60.0027
2	82.3421	76.2342	75.1137	67.2311	65.0033	60.0456
3	83.0302	76.7802	76.0232	67.6121	67.9111	60.0101
4	83.0012	76.8767	75.3545	67.5689	65.5889	60.01
5	82.4632	76.2069	75.1182	66.7865	65.3421	60.0043
6	81.8795	76.1951	74.1147	66.0011	65.001	59.9876
7	80.789	76.0022	73.2433	65.5045	64.7768	59.3425
8	79.6574	75.5467	72.1389	64.8675	64.3345	59.0001
9	78.6759	74.3452	71.3511	63.7865	62.9976	58.3456

下載: 導出CSV

表 5 部分樣品薄片鑒定結果

Table 5. Identification results of some sample slices

Sample number	Sampling location		Lithology	Mineral composition
Sample number	Longitude	Latitude	Lithology	Mineral composition
D0158	96°32′13.1″	41°59′41.4″	Albite epidote chlorite schist	Quartz 48%，Feldspar 25%，Chlorite 12%，Epidote 8%， Biotite 3%，Sphene 3%
D0126	96°32′53.1″	41°59′20.4″	Plagioclase amphibolite	Amphibole 75%，Plagioclase 15%，Quartz 5%，Carbonate minerals 4%，Opaque minerals 1%
D0123	96°32′56.6″	41°59′19.5″	Garnet muscovite plagioclase gneiss	Muscovite 30%，Feldspar 62%，Quartz 5%，Garnet 3%
D0121	96°32′57.6″	41°59′19.3″	Biotite plagioclase gneiss	Quartz 45%，Plagioclase 33%，Biotite 20%，Amphibole 2%
D0117	96°33′8.2″	41°59′17.8″	Biotite plagioclase gneiss	Quartz 28%，Plagioclase 25%，Biotite 25%，Chlorite 15%， K-feldspar 5%，Opaque minerals 2%
D0107	96°33′16.8″	41°59′7.5″	Tonalite	Quartz 42%，Feldspar 34%，Biotite 12%，Amphibole 8%，Carbonate minerals 2%，Sericite 2%
D0105	96°33′16.6″	41°59′4″	Garnet biotite plagioclase gneiss	Quartz 50%，Feldspar 31%，Biotite 10%，Garnet 5%， Opaque minerals 3%，Apatite 1%
D0103	96°33′16.3″	41°59′3″	Garnet biotite plagioclase gneiss	Quartz 40%，Plagioclase 30%，K-feldspar 7%，Biotite 12%，Garnet 3%
D1125	96°31′36.8″	41°57′24.3″	Serpentine diopside dolomitic marble	Calcite 49%，Diopside 32%，Serpentine 14%，Actinolite 5%
D0901	96°40′47.7″	41°55′39.2″	Albite chlorite quartz phyllite	Quartz 32%，Plagioclase 16%，Chlorite 48%，Epidote 4%，Sphene
D0420	96°34′35.6″	41°52′36.5″	Plagioclase amphibolite	Plagioclase 55%，Amphibole 30%，Quartz 10%，Biotite 3%，Epidote 2%
D0426	96°34′34.5″	41°52′29″	Mica quartz schist	Quartz 50%，Muscovite 30%，Biotite 20%
D0450	96°34′14.9″	41°51′38.6″	Plagioclase hornblende gneiss	Plagioclase 35%，Amphibole 55%，Epidote 5%，Quartz 5%
D1007	96°30′44.9″	41°50′45.1″	Mica monzonite gneiss	K-feldspar 30%，Plagioclase 25%，Quartz 27%， Biotite 14%，Muscovite 4%
D0055	96°33′49″	41°50′6.3″	Biotite plagioclase gneiss	Plagioclase 38%，Quartz 35%，Biotite 20%，Muscovite 3%，Chlorite 4%
D0113	96°34′19.3″	41°51′46″	Plagioclase hornblende gneiss	Amphibole 45%，Albite 37%，Microcline 10%，Quartz 5%，Pyroxene 2%，Opaque minerals 1%

下載: 導出CSV

表 6 精度評價

Table 6. Accuracy evaluation

Mineral	PA/%	UA/%	OA/%	Kappa
Bi	80.95	78.64	85.4088	0.7779
Mus	83.35	75.60	84.7640	0.7833
Am	79.74	77.55	85.7308	0.7748
Chl	68.74	63.76	70.6933	0.5938
Gt	58.42	58.43	65.5992	0.5462
Act	59.61	64.41	66.7509	0.5560

下載: 導出CSV

表 7 本文方法與其他方法提取精度對比

Table 7. Comparison of extraction accuracy between the present method and other methods

Mineral	Extraction accuracy/%
Mineral	Ratio + threshold segmentation	Ratio +SVM	Method of this paper
Bi	77.3546	81.1341	85.4088
Mus	81.2830	82.9870	84.7640
Am	76.9684	81.1002	85.7308
Chl	64.6594	66.4532	70.6933
Gt	61.8422	63.1886	65.5992
Act	62.7847	64.2351	66.7509

下載: 導出CSV

www.77susu.com

參考文獻(22)

[1]	Diener J F A, White R W, Link K, et al. Clockwise, low-P metamorphism of the Aus qranulite terrain, southern Namibia, during the Mesoproterozoic Namaqua Oroqeny. Precambrian Res, 2013, 224: 629 doi: 10.1016/j.precamres.2012.11.009
[2]	Xie M H, Zhang Q, Chen S B, et al. Extraction of alteration anomaly information by feature-based principal component analysis from ASTER data. Editorial Committe Earth Sci J China Univ Geosci, 2015, 40(8): 1381 謝明輝, 張奇, 陳圣波, 等. 基于特征導向主成分分析遙感蝕變異常提取方法. 地球科學—中國地質大學學報, 2015, 40(8):1381
[3]	Zadeh M H, Tangestani M H, Roldan F V, et al. Sub-pixel mineral mapping of a porphyry copper belt using EO-1 Hyperion data. Adv Space Res, 2014, 53(3): 440 doi: 10.1016/j.asr.2013.11.029
[4]	Wu Z C, Ye F W, Guo F S, et al. A review on application of techniques of principle component analysis on extracting alteration information of remote sensing. J Geo-Inf Sci, 2018, 20(11): 1644 doi: 10.12082/dqxxkx.2018.180195 吳志春, 葉發旺, 郭福生, 等. 主成分分析技術在遙感蝕變信息提取中的應用研究綜述. 地球信息科學學報, 2018, 20(11):1644 doi: 10.12082/dqxxkx.2018.180195
[5]	Liu Y Z, Lai H R, Zhang D W, et al. Change detection of high resolution remote sensing image alteration based on multi-feature mixed kernel SVM model. Remote Sens Land Resour, 2019, 31(1): 16 劉義志, 賴華榮, 張丁旺, 等. 多特征混合核 SVM 模型的遙感影像變化檢測. 國土資源遙感, 2019, 31(1):16
[6]	He Z H, He B B. Weight spectral angle mapper (WSAM) method for hyperspectral mineral mapping. Spectrosc Spectr Anal, 2011, 31(8): 2200 doi: 10.3964/j.issn.1000-0593(2011)08-2200-05 何中海, 何彬彬. 基于權重光譜角制圖的高光譜礦物填圖方法. 光譜學與光譜分析, 2011, 31(8):2200 doi: 10.3964/j.issn.1000-0593(2011)08-2200-05
[7]	Kaur S, Bansal R K, Mittal M, et al. Mixed pixel decomposition based on extended fuzzy clustering for single spectral value remote sensing images. J Indian Soc Remote Sens, 2019, 47(3): 427 doi: 10.1007/s12524-019-00946-2
[8]	Feng W Q, Sui H G, Tu J H, et al. Change detection method for high resolution remote sensing images using random forest. Acta Geodaet Cartograph Sin, 2017, 46(11): 1880 doi: 10.11947/j.AGCS.2017.20170074 馮文卿, 眭海剛, 涂繼輝, 等. 高分辨率遙感影像的隨機森林變化檢測方法. 測繪學報, 2017, 46(11):1880 doi: 10.11947/j.AGCS.2017.20170074
[9]	Booysen R, Zimmermann R, Lorenz S, et al. Towards multiscale and multisource remote sensing mineral exploration using RPAS: a case study in the Lofdal carbonatite-hosted REE deposit, Namibia. Remote Sens, 2019, 11(21): 2500 doi: 10.3390/rs11212500
[10]	Cid Y D, Muller H, Platon A, et al. 3D solid texture classification using locally-oriented wavelet transforms. IEEE Trans Image Process, 2017, 26(4): 1899 doi: 10.1109/TIP.2017.2665041
[11]	Wang M, Zhang X C, Wang J Y, et al. Forest resource classification based on random forest and object oriented method. Acta Geodaet Cartograph Sin, 2020, 49(2): 235 doi: 10.11947/j.AGCS.2020.20190272 王猛, 張新長, 王家耀, 等. 結合隨機森林面向對象的森林資源分類. 測繪學報, 2020, 49(2):235 doi: 10.11947/j.AGCS.2020.20190272
[12]	You Y F, Wang S Y, Wang B, et al. Study on hierarchical building extraction from high resolution remote sensing imagery. J Remote Sens, 2019, 23(1): 125 游永發, 王思遠, 王斌, 等. 高分辨率遙感影像建筑物分級提取. 遙感學報, 2019, 23(1):125
[13]	Li J Y, Zhao Y K, Xue Z E, et al. A survey of model compression for deep neural networks. Chin J Eng, 2019, 41(10): 1229 李江昀, 趙義凱, 薛卓爾, 等. 深度神經網絡模型壓縮綜述. 工程科學學報, 2019, 41(10):1229
[14]	Cracknell M J, Reading A M. Geological mapping using remote sensing data: A comparison of five machine learning algorithms, their response to variations in the spatial distribution of training data and the use of explicit spatial information. Comput Geosci, 2014, 63: 22 doi: 10.1016/j.cageo.2013.10.008
[15]	Harris J. R, He J X, Rainbird R, et al. A comparison of different remotely sensed data for classifying bedrock types in Canada’s arctic: application of the robust classification method and random forests. Geosci Can, 2014, 41(4): 557
[16]	Hossain M D, Chen D M. Segmentation for object-based image analysis (OBIA): A review of algorithms and challenges from remote sensing perspective. ISPRS J Photogramm Remote Sens, 2019, 150: 115 doi: 10.1016/j.isprsjprs.2019.02.009
[17]	Diaz G F, Ortiz J M, Silva J F, et al. Variogram-based descriptors for comparison and classification of rock texture images. Math Geosci, 2020, 52(4): 451 doi: 10.1007/s11004-019-09833-5
[18]	Zhang L, Liu Z, Ren T W, et al. Identification of seed maize fields with high spatial resolution and multiple spectral remote sensing using random forest classifier. Remote Sens, 2020, 12(3): 362 doi: 10.3390/rs12030362
[19]	Zhu J J, Fan X T, Du X P. Geometric feature representation and building extraction based on geometric features. J Appl Sci, 2015, 33(1): 9 doi: 10.3969/j.issn.0255-8297.2015.01.002 朱俊杰, 范湘濤, 杜小平. 幾何特征表達及基于幾何特征的建筑物提取. 應用科學學報, 2015, 33(1):9 doi: 10.3969/j.issn.0255-8297.2015.01.002
[20]	Masoumi F, Eslamkish T, Abkar A A, et al. Integration of spectral, thermal, and textural features of ASTER data using random forests classification for lithological mapping. J Afric Earth Sci, 2017, 129: 445 doi: 10.1016/j.jafrearsci.2017.01.028
[21]	Pournamdari M, Hashim M, Pour A B. Spectral transformation of ASTER and Landsat TM bands for lithological mapping of Soghan ophiolite complex, South Iran. Adv Space Res, 2014, 54(4): 694 doi: 10.1016/j.asr.2014.04.022
[22]	Zhang B, He B B. Multi-scale segmentation of high-resolution remote sensing image based on improved watershed transformation. J Geo-Inf Sci, 2014, 16(1): 142 張博, 何彬彬. 改進的分水嶺變換算法在高分辨率遙感影像多尺度分割中的應用. 地球信息科學學報, 2014, 16(1):142