基于NDSI和NDISI指数的SPOT-5影像裸土信息提取

遥感裸土识别制图为水土流失治理工作提供了科学依据。本文以SPOT-5影像为实验数据,提出一种以土壤指数NDSI和不透水面指数NDISI提取裸土的方法。通过热红外波段的亚像元分解技术,将同期120 m分辨率的TM 6波段细化为10 m分辨率的地表温度影像,为SPOT-5影像计算NDISI不透水面指数增加了必要的热红外波段。在此基础上,构建双重指数模型,获得10 m分辨率的裸土数据。研究表明,双重指数模型可较好地解决裸土提取中建筑用地与裸土相混淆的问题,提取裸土的总精度可达95.4%。通过比较10 m的SPOT-5和30 m的TM影像的裸土提取结果,发现影像分辨率的提升可使裸土信息提取结果更加准确、精细。因此,本文为更高分辨率裸土识别制图,提供了一种有效的方法。

Extraction of Bare Soil Features from SPOT-5 Imagery Based on NDSI and NDISI

The accurate mapping of bare soil land is of vital significance to soil erosion treatment. Satellite remote sensing has become a popular technology used in bare soil mapping, owing to the capability of monitoring bare soil dynamics in a cost-effective manner. Nowadays, Landsat imagery with a spatial resolution of 30 m has been widely used in previous work for bare soil extraction. Nevertheless, due to the limitation of its spatial resolution, the Landsat image can hardly meet the requirement of fine-scale bare soil mapping. Thus, a method for higher resolution bare soil mapping was proposed in this paper using SPOT-5 imagery as the main data. The Normalized Difference Soil Index (NDSI) was firstly utilized to enhance the bare soil features from the SPOT-5multispectral 10 m image. However, the enhanced soil information was mixed with the built-up land features due to the confusion between the two categories. To solve this problem, the Normalized Difference Impervious Surface Index (NDISI) was further introduced to separate the built-up features from the bare soils. Due to the lack of a thermal band in SPOT-5 imagery, which is required in computing NDISI, a near synchronous Landsat TM thermal band 6 was fine-sharpened to 10 m resolution, and afterwards, it was added to the SPOT-5 bands to calculate NDISI. Accordingly, a model was used in mapping the bare soil, based on the NDSI and NDISI. Finally, the extraction results of bare soil at 10 m resolution were obtained through the model, which achieved an overall accuracy of 95.4%. It suggests that the two-index based method can effectively eli- minate the confusion between bare soil land and built-up land, and can extract the bare soil features with a higher accuracy. In addition, the bare soil features extracted from SPOT imagery and TM imagery respectively, were further compared. It is found that a higher spatial resolution can lead to a more accurate extraction result. Therefore, the 10 m resolution SPOT-5 image is considered to be more adaptive and useful in bare soil mapping for soil erosion treatment.