论文部分内容阅读
使用高空间分辨率卫星WorldView-2的多光谱遥感影像,构建植被指数和纹理因子等遥感因子与森林地上生物量的关系方程,并计算模型估测精度和均方根误差,探索高分辨率数据的光谱与纹理信息在温带森林地上生物量估测应用中的潜力。以黑龙江省凉水自然保护区温带天然林及天然次生林为研究对象,通过灰度共生矩阵(GLCM)、灰度差分向量(GLDV)及和差直方图(SADH)对高分辨率遥感影像进行纹理信息提取,并利用外业调查的74个样地地上生物量与遥感因子建立参数估计模型。提取的遥感因子包括6种植被指数(比值植被指数RVI、差值植被指数DVI、规一化植被指数NDVI、增强植被指数EVI、土壤调节植被指数SAVI和修正的土壤调节植被指数MSAVI)以及3类纹理因子(GLCM、GLDV和SADH)。为避免特征变量个数较多对估测模型造成过拟合,利用随机森林算法对提取的遥感因子进行特征选择,将最优的特征变量输入模型参与建模估测。采用支持向量回归(SVR)进行生物量建模及验证,结果显示选入模型的和差直方图均值(sadh_mean)、灰度共生矩阵方差(glcm_var)和差值植被指数(DVI)等遥感因子对森林地上生物量有较好的解释效果;植被指数+纹理因子组合的模型获得较精确的AGB估算结果(R2=0.85,RMSE=42.30 t/ha),单独使用植被指数的模型精度则较低(R~2=0.69,RMSE=61.13 t/ha)。
Using the multispectral remote sensing image of WorldView-2, a satellite with high spatial resolution, the relationship equation between vegetation remote sensing factors, such as vegetation index and texture factor, and aboveground biomass was constructed and the model estimation accuracy and root mean square error were calculated to explore high resolution data The potential of spectral and texture information for biomass estimation in temperate forests. Taking the temperate natural forest and the natural secondary forest in Liangshui Nature Reserve of Heilongjiang Province as the research object, the texture information of high resolution remote sensing image was analyzed by GLCM, GLDV and SADH. Extraction, and using the field survey of 74 plots of aboveground biomass and remote sensing parameters to establish a parameter estimation model. The extracted remote sensing factors include six kinds of vegetation index (RVI, DVI, NDVI, EVI, SAVI and MSAVI) and three types of vegetation indices Texture Factors (GLCM, GLDV and SADH). In order to avoid over fitting of the estimation model with a large number of feature variables, a random forest algorithm is used to select the extracted remote sensing factors, and the optimal feature variables are input into the model for modeling and estimation. The biomass modeling and validation using Support Vector Regression (SVR) showed that remote sensing parameters such as sadh_mean, glcm_var and DVI were selected for the model. The aboveground biomass of forest had a better interpretation effect. The model of vegetation index + texture factor obtained more accurate AGB estimation (R2 = 0.85, RMSE = 42.30 t / ha), and the model accuracy of vegetation index alone was lower R ~ 2 = 0.69, RMSE = 61.13 t / ha).