摘要： 露天开采会彻底改变原有土地利用景观格局，直接破坏当地生态环境，甚至会影响附近居民的生产和生活，因此越来越多的学者开始关注开采扰动。先前有关利用时序多光谱影像提取开采扰动的研究区集中于扰动形式单一的森林区。而我国露天煤矿大多集中于草原区，且我国东北部的草原矿区因其脆弱的生态环境以及其他多种扰动形式的存在，使得开采扰动识别更加困难。为明确我国东北部生态脆弱区草原露采场的开采扰动，以胜利矿区为例，利用1986年—2017年27期Landsat多光谱遥感影像，基于归一化植被指数NDVI（normalized difference vegetation index）的长时间序列轨迹变化特征（为了去除物候、云和阴影等对时序多光谱影像的影响，利用BISE-WT滤波器对原始NDVI时间序列进行滤波处理, 有效地去除时序NDVI数据中的噪声并同时保留有效信息），经过样本点训练，获得CV阈值（变异系数coefficient of variation）和Max阈值（植被阈值），构建CV-Max扰动识别模型，提取研究区的扰动分布。并利用植被阈值，分析NDVI时序轨迹，获得扰动年际信息，重构扰动历史地图；进而通过分析研究区典型地物的光谱特征，构建裸煤提取规则，以此来提取研究区的裸煤分布；最后通过构建裸煤及扰动区两者间的拓扑关系，进行空间拓扑叠置分析，从而获得开采扰动信息。经过精度验证，开采扰动的提取精度达到93.17%（Kappa系数=0.85），扰动年际信息提取精度达到83.35%（Kappa系数=0.81）。结果表明：在研究期间，空间上，开采扰动面积占研究区总面积的8.90%；时间上，开采扰动的发生集中于2000年—2009年，期间开采扰动像元占开采扰动总像元的76.70%；1988年—1998年矿区属于土地损毁初始期，2000年—2005年矿区属于土地损毁加速期，2006年—2009年矿区属于土地损毁高峰期，2010年—2017年开采扰动像元占比趋势比较平缓且持续处于较低水平，矿区土地损毁范围基本稳定。所提出的针对我国东北部生态脆弱性草原矿区，基于时序多光谱影像，利用植被指数NDVI和裸煤光谱特征提取开采扰动信息的方法是可行的，该研究结果可为干旱、半干旱草原露天矿区的可持续发展提供数据和理论方法支撑。
关键词：Landsat时序影像；NDVI；光谱特征；开采扰动；CV-Max模型
Abstract：Surface mining will completely change the original landscape pattern of land use, directly destroy the local ecological environment, and even affect the production and life of the nearby residents; therefore, more and more scholars have begun to pay attention to mining disturbance. Previous studies on extracting mining disturbances from temporal multispectral images focused on forest areas with single disturbance form. However, most surface mines in China are concentrated in grassland areas, and in the grassland mining areas in northeastern China, mining disturbances are more difficult to be identified because of their fragile ecological environment and the existence of various other forms of disturbance. In order to clarify the mining disturbance of grassland open stope in ecologically fragile areas in northeastern China, the authors taking Shengli mining area as an example, firstly employs 27 Landsat multi-spectral remote sensing images from 1986 to 2017, and bases the study on the long time series trajectory change characteristics of NDVI (normalized difference vegetation index). (In order to remove the effects of phenology, cloud and shadow on time series multispectral images, BISE-WT filter is used to filter the original NDVI time series to effectively remove the noise in the time series NDVI data and retain the effective information at the same time). After sample point training, CV threshold (Coefficient of Variation) and Max vegetation threshold are obtained. The Max vegetation threshold (vegetation threshold) is then used to construct the CV-Max disturbance recognition model and extract the disturbance distribution in the study area. Furthermore, using vegetation threshold, NDVI time series trajectory is analyzed to obtain disturbance interannual information and reconstruct disturbance history map. Then, by analyzing the spectral characteristics of typical terrain in the study area, bare coal extraction rules are constructed to extract the distribution of bare coal in the study area. Finally, the topological relationship between bare coal and disturbance area is constructed and a spatial topological overlay analysis is conducted to obtain mining disturbance information. The accuracy verification reveals the extraction accuracy of mining disturbance is 93.17% (Kappa coefficient=0.85) and the extraction accuracy of disturbance interannual information is 83.35% (Kappa coefficient=0.81) respectively. The results show that during the study period, the mining disturbance area accounts for 8.90% of the total area of the study area in space; in terms of time, the occurrence of mining disturbance concentrated in 2000—2009, during which the mining disturbance pixels accounted for 76.70% of the total mining disturbance pixels; the years from 1988 to 1998 witness the initial period of land destruction, and in 2000—2005, land destruction increased in the mining area, and in 2006—2009, the land destruction the mining area reached the peak. The proportion trend of mining disturbance pixels in 2010—2017 is relatively flat and continues to be at a low level, and the scope of land damage in mining area is basically stable. In view of the ecologically fragile grassland mining area in northeastern China, the method of extracting mining disturbance information by using NDVI and bare coal spectral features based on time series multispectral images is feasible. The research results can provide data and theoretical method support for the sustainable development of arid and semi-arid grassland surface mining area.
Key words：Landsat multi-temporal; NDVI; Spectral features; Mining disturbance; CV-Max model
李 晶，邓晓娟，杨 震，刘乾龙，王 媛，崔绿园. 基于时序多光谱影像的干旱草原区开采扰动信息提取方法[J]. 光谱学与光谱分析, 2019, 39(12): 3788-3793.
LI Jing, DENG Xiao-juan, YANG Zhen, LIU Qian-long, WANG Yuan, CUI Lü-yuan. A Method of Extracting Mining Disturbance in Arid Grassland Based on Time Series Multispectral Images. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2019, 39(12): 3788-3793.
[1] Michael S, Richard L, Peter B, et al. Remote Sensing of Environment, 2015, 158(1)：156.
[2] Huang C B, Dian Y Y, Zhou Z X, et al. Journal of Remote Sensing, 2015, 19(4)：657.
[3] Kennedy R E, Yang Z Q, Cohen W B. Remote Sensing of Environment, 2010, 114(12): 2897.
[4] Verbesselt J, Hyndman R, Newnham G, et al. Remote Sensing of Environment, 2010, 114(1): 106.
[5] Verbesselt J, Hyndman R, Zeileis A, et al. Remote Sensing of Environment, 2010, 114(12): 2970.
[6] Verbesselt J, Zeileis A, Herold M. Remote Sensing of Environment, 2012, 123: 98.
[7] Yang C, Shen R P, Yu D W, et al. Journal of Remote Sensing, 2013, 17(5): 1246.
[8] Huang C, Goward S N, Masek J G, et al. Remote Sensing of Environment, 2010, 114(1)：183.
[9] Zhu Z, Woodcock C E, Olofsson P. Remote Sensing of Environment, 2012, 122: 75.
[10] Sen S, Zipper C E, Wynne R H, et al. Photogrammetric Engineering and Remote Sensing, 2012, 78(3): 223.
[11] Li J, Zipper C E, Donovan P F, et al. Environ. Monit. Assess, 2015, 187(9)：557.
[12] Yang Z, Li J, Shen Y Y, et al. International Journal of Remote Sensing, 2018, 39(12): 3816.
[13] Yang Z, Li J, Zipper Carl E, et al. Science of the Total Environment, 2018, 644(10): 916.
[14] Yang Z, Li J, Yin S Q, et al. Remote Sensing Letters, 2018, 9(12): 1224.
[15] LI Jing, Zipper C E, LI Song, et al(李 晶, Zipper Carl E, 李 松, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2015, 31(16): 251.
[16] LI Jing, JIAO Li-peng, SHEN Ying-ying, et al(李 晶，焦利鹏，申莹莹，等). Journal of China Coal Society(煤炭学报), 2016, 41(11): 2822.
