img

官方微信

高级检索

中国沙漠 ›› 2006, Vol. 26 ›› Issue (6): 926-931.

• 沙漠与沙漠化 • 上一篇    下一篇

基于地质统计学影像纹理的海南沙漠化监测研究

廖楚江, 王长耀, 林文鹏   

  1. 中国科学院遥感应用研究所 遥感科学国家重点实验室, 北京 100101
  • 收稿日期:2005-07-04 修回日期:2005-09-05 出版日期:2006-12-20 发布日期:2006-12-20

Research on Desertification Monitoring for Hainan Province based on Geostatistical Texture

LIAO Chu-jiang, WANG Chang-yao, LIN Wen-peng   

  1. The State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing Applications, Chinese Academy of Sciences, Beijing 100101, China
  • Received:2005-07-04 Revised:2005-09-05 Online:2006-12-20 Published:2006-12-20

摘要: 海南省东部地区在沙漠化监测上呈现两大难点:一是海滩沙地与沙漠化土地在影像上呈现近乎相同的光谱特征,基于传统的遥感影像光谱分类方法无法得到实际的沙漠化面积;二是该地区属于热带沿海地区,常规的监测指标体系与实际情况相去甚远,必须寻求其他的手段来进行沙漠化程度的分级。基于不同沙地类型在地表空间结构上的差异,本文提出将地质统计学纹理方法应用到沙漠化监测中,通过变异函数纹理来加大各种不同类别沙地间的区别,提高样本选择的分离度。结果表明,运用变异函数纹理结合光谱波段的最大似然分类方法能够很好地界定海滩沙地和沙漠化土地的不同等级,依据分类结果计算得到的沙漠化土地面积与统计数据吻合较好,总精度达到92.4%,证明了地质统计学纹理在实现该地区遥感沙漠化监测方面的有效性,同时也为其他地区沙漠化监测找到一个可资借鉴的方法。

关键词: 地质统计学, 纹理, 变异函数, 沙漠化, 沙地

Abstract: During the research on desertification for Wencang, Hainan Province, we found that there are two difficulties in monitoring desertification of this area comparing to the research about desertification of Northwest China. One is that the beach sandlot shows same spectral features with the inland desert, so, to use traditional spectral classification to discriminate them is impossible, accordingly we can't get the real desertification area; the other is that the desertification degree in this area is also difficult to grade directly from origin radiometric bands, and because current desertification area is mainly formed by excessively destroying, thus, the experiences for Northwest China desertification monitoring are not fit, we must find other measures to grade desertification degree. Based on spatial construction differences of sandlot types, the author firstly put forward using geostatistical texture to monitor desertification, using the variogram texture to increase the discrimination degree of different samples. The result suggested that the maximum likelihood classification based on the combination of variogram texture and spectral bands can perfectly partition the beach sandlot and the different grades of inland desertification, which prove that geostatistical texture is validate in the application of desertification monitoring.

Key words: Geostatistics, texture, variogram, desertification, sandlot

中图分类号: