基于三指数合成影像的西北地区城市建筑用地 遥感信息提取研究

随着西部大开发战略的实施以及“一带一路”战略的影响,西北地区的城市发展也发生着巨大变化,利用遥感影像更加准确地提取西北地区城市建筑用地信息对分析城市扩张趋势、规划城市建设具有重要意义。本文以2000年兰州市主城区和2003年西宁市主城区的Landsat 7 ETM ⁺影像为数据源,结合压缩数据维的方法,通过构建三指数合成影像并利用该影像来提取城市建筑用地信息。实验首先根据兰州市主城区的影像光谱特征,创建了归一化差值裸地指数（NDBLI）。然后将该指数与比值居民地指数（RRI）、修正型归一化水体指数（MNDWI）合成为一个包含3个波段的新型三指数合成影像NRM（NDBLI、RRI、MNDWI）;同时,根据集成学习思想,为增强城市建筑用地信息,将主成分分析的第一波段（PC1）、归一化差值建筑用地指数（NDBI）和比值居民地指数（RRI）合成为一个包含3个波段的新型三指数合成影像PNR（PC1、NDBI、RRI）;最后分别将三指数合成影像NRM和三指数合成影像PNR作最大似然分类提取城市建筑用地信息,将其提取结果与由归一化差值建筑用地指数（NDBI）、修正型归一化水体指数（MNDWI）和土壤调节植被指数（SAVI）所创建的NMS（NDBI、MNDWI 、SAVI）影像得到的最大似然分类结果作精度比较,并利用西宁市主城区影像对本文方法进行了相应验证。结果表明,利用三指数合成影像PNR提取城市建筑用地的总精度和Kappa系数最高,其总精度达到了90%以上,适合于提取西北地区含裸地较多的城市建筑用地。

Mapping Urban Built-Up Land in Northwest China based on Three-Index Synthetic Remote Sensing Imagery

With the implementation of the Great Western Development Strategy and the Belt-Road Initiative, urban development in northwest China has undergone tremendous changes. Extracting built-up land information in northwest China more accurately by remote sensing imagery is important for sustainable urban planning. By using Landsat TM/ETM ⁺ imagery, this paper proposed a new method for extracting built-up land information in the urban districts in Lanzhou City and Xining City. By detailed spectral signature analysis of the image, the Normalized Difference Bare Land Index (NDBLI) was calculated, which was then synthesized with the Ratio of Residential area index (RRI) and the Modified Normalized Water Index (MNDWI). In so doing, a new type of three-index synthetic image NRM (NDBLI, RRI, MNDWI) was generated. Meanwhile, following the idea of ensemble learning, the first band of principal component analysis(PC1), the Normalized Difference Built-up Index (NDBI), and the Ratio of Residential area Index (RRI) were synthesized to enhance urban built-up land information, which formed another three-index synthetic image PNR (PC1, NDBI, RRI). Then, the maximum likelihood based supervised classification was performed on the new synthesized image NRM (NDBLI, RRI, MNDWI) and the new synthesized image PNR (PC1, NDBI, RRI). Finally, the urban built-up land was extracted by masking out non-built-up land classes. These classification results were compared with the results of the maximum likelihood classification based on the three-index synthetic image NMS (NDBI, MNDWI, SAVI) synthesized by the three indexes of the Normalized Difference Built-up Index (NDBI), the Modified Normalized Water Index (MNDWI), and the Soil Adjustment Vegetation Index (SAVI). Under different spatiotemporal conditions, the three-index synthetic image NRM (NDBLI, RRI, MNDWI), three-index synthetic image PNR (PC1, NDBI, RRI), and three-index synthetic image NMS (NDBI, MNDWI, SAVI) were used to extract the urban build-up land information of Lanzhou City and Xining City. Results show that the overall accuracy and kappa coefficient of classification using the three-index synthetic image PNR (PC1, NDBI, RRI) were higher than those based on the three-index synthetic image NRM (NDBLI, RRI, MNDWI) and three-index synthetic image NMS (NDBI, MNDWI, SAVI). The urban built-up land information was thus extracted by the three-index synthetic image PNR (PC1, NDBI, RRI), with the overall accuracy above 90%. Our findings suggest that the three-index synthetic image PNR (PC1, NDBI, RRI) is suitable for extracting urban built-up land in northwest China.