论文部分内容阅读
选择线性混合像元分解模型、亚像元模型、最大三波段梯度差法模型以及修正的三波段梯度差法的2个变异模型来提取植被覆盖度,结合地面实测数据,探讨了提取干旱区荒漠稀疏植被覆盖度信息的适宜模型,并以简单平均法模拟了不同尺度的覆盖度影像,通过尺度上推检验了模型在MODIS尺度上的反演效应.结果表明:线性混合像元分解模型反演覆盖度的精度高于其他模型,适于稀疏植被地区,但端元的正确选取较难,从而影响其运用;亚像元分解模型是一个通用模型,植被分类图越精细,通过亚像元分解模型得到的覆盖度精度越高,但这也同时意味着该模型需要测定大量的输入参数;最大三波段梯度差法的算法简单、易于操作,其在农田等中高植被覆盖区及裸土区的预测值与实测值接近,但对干旱区稀疏植被的估计精度偏低;修正后的三波段最大梯度差法模型在稀疏植被覆盖区的预测值与实测值基本一致,在不同尺度上反演的覆盖度信息与实测值的一致性较好.该方法可有效提取干旱区低覆盖度植被信息.
Two linear regression models of pixel decomposition, sub-pixel model, maximum three-band gradient difference method and modified three-band gradient method were selected to extract vegetation coverage. Combined with the measured data, The model of sparse vegetation coverage is modeled, and the images of coverage at different scales are modeled by the simple average method, and the inversion effect on the MODIS scale is tested by scaling up.The results show that the inversion of the linear mixed pixel decomposition model The accuracy of coverage is higher than other models, which is suitable for sparse vegetation areas. However, the correct selection of end elements is more difficult and affects their application. The sub-pixel decomposition model is a general model, and the finer the vegetation classification maps, The accuracy of the coverage obtained by the model is higher, but this also means that the model needs to measure a large number of input parameters. The algorithm of the maximum three-band gradient difference method is simple and easy to operate. The predicted value is close to the measured value, but the estimation precision of the sparse vegetation in the arid area is low. The modified three-band maximum gradient difference method is more effective in sparse vegetation VTA predicted and measured values is consistent, the consistency of coverage information and the measured value of the inversion at different scales better. This method can extract information arid low vegetation coverage area.