【摘要】 以地面样点为基础的森林自然度评价方法很难获得区域范围森林自然度等级,针对该问题,提出了利用高分遥感卫星影像数据,划分区域范围森林自然度等级的方法。以湖北竹山县九华山林场为试验区域,在选取研究区典型样地的基础上,结合高分二号(GF-2)遥感影像数据的特点,从GF-2影像上提取遥感光谱、纹理等特征并结合地形特征,采用随机森林算法在大尺度范围对九华山林场森林自然度等级进行分类研究。结果发现:以GF-2数据为基础提取的植被指数、光谱、纹理等特征与地形特征结合,采用随机森林算法可较好地划分森林自然度等级,总体分类精度高达93.97%, Kappa系数为0.91。对森林自然度等级影响最重要的6个特征因子为高程、坡向、坡度、纹理均值、光谱主成分变化分量和归一化植被指数(NDVI)。结果表明,基于遥感影像提取的特征和地形特征结合进行森林自然度等级划分的研究方法具有可行性,为大面积区域的森林自然度等级划分奠定基础。更多还原

【Abstract】 It is difficult to obtain forest naturalness levels at a regional scope based on in situ plots. In response to this problem, this study proposed a method to classify the naturalness levels of forests in regional areas through high spatial resolution remote sensing image data gathered from research conducted in Jiuhuashan Forest Farm in Zhushan County, Hubei Province. We obtained different typical plots and high spatial resolution Gaofen-2(GF-2) remote sensing image data in the study area; spectrum, texture, and vegetation indices were extracted from the GF-2 images. Combining the features extracted from these images, as well as topographic features, namely, elevation, slope, and aspect, a random forest algorithm was used to classify the forest naturalness level of Jiuhuashan Forest. The results showed that the classification of the vegetation, spectrum, and texture indices and a combination of the terrain features with a random forest algorithm based on GF-2 remote sensing data can better classify the naturalness of the forest;the overall classification accuracy reached 93.97% and the Kappa coefficient was 0.91. The six most important factors affecting the naturalness level of the forest were elevation, aspect, slope, texture mean, the spectral principal component change, and the normalized difference vegetation index(NDVI). In this manner, we confirm that this combination of factors and methods can help in classifying forest′s natural degrees, laying a foundation for the natural degree division of forests in large areas.更多还原