基于纹理的高寒地区人为扰动地表信息提取

Extraction of Artificially Disturbed Surface Information Based on Texture Feature in Alpine Region

高寒地区气候寒冷、生态环境脆弱。近年来,频繁的人类工程活动严重破坏地表植被,加剧了高寒地区的土壤侵蚀。为有效监管生产建设活动,遏制人为新增土壤侵蚀,开展及时、准确的监测分析迫在眉睫,探索符合高寒区域的人为扰动地表信息提取方法十分必要。目前仅靠单波段灰度图像纹理特征提取地表信息存在信息量不足等明显缺陷,改进的彩色图像灰度共生矩阵方法(Color gray-level co-occurrence matrix,简称CGLCM),可为人为扰动地表识别提供更为精确的纹理特征信息。本文以西藏自治区墨竹工卡县甲玛区为研究区,基于CGLCM提取出纹理特征并结合植被指数(NDVI)数据与光谱数据,运用面向对象分类方法实现人为扰动地表信息提取。结果表明,CGLCM方法相较于NDVI方法精度提高了5.12%,达到90.88%;Kappa系数提高了0.07,为0.87,结合纹理特征对影像分类可有效提高分类精度。据此,本文基于纹理特征的扰动地表信息提取方法为高寒地区提供了可靠的人为扰动信息提取途径,实现了基于遥感技术的高寒地区扰动地表信息快速自动监测。

The climate is cold and the ecological environment is fragile in the alpine region. In recent years,the frequent human engineering activities destroy the surface vegetation seriously which aggravate the soil erosion. In order to supervise the production-construction activities and curb the new soil erosion effectively,it is urgent to do some work about monitoring analysis accurately,and explore the method of extracting human disturbance information which meets the characteristics of alpine region. There are some weaknesses that it is lack of enough information with the extraction method of a single band grayscale image texture feature at present. However,the method of improved color image gray co-occurrence matrix（ Color gray-level co-occurrence matrix,referred to as CGLCM）which can provide more accurate texture feature information for human disturbance monitoring classification. We chose the study area which is a part of Mozhugongka County of Jiama area in Tibet Province extracting the information of artificial disturbance area by the method of object-oriented classification with the combination of the texture feature based on CGLCM,vegetation index（ NDVI） data and spectral data. The result of the experiments shows that the accuracy of extraction is improves of 5.12%,which reaches 90.88% by the method of CGLCM,and the kappa coefficient is 0. 87,increases by 0. 07 compared with the NDVI method. Image classification based on texture feature can improve the classification accuracy effectively. Therefore,the paper provides a reliable approach based on texture features to extract the information of human disturbance,and realizes fast automatic monitoring of disturbing surface information in alpine region based on Remote Sensing.