基于时序光谱特征匹配的中分辨率遥感橡胶林提取

摘要

Texture and phenology was the effective diagnostic features to distinguish the rubber plantation from natural forests and other land types.Extracted the rubber plantation endmember from the multispectral image which synthesized with time series images, used the spectral angle and the spectral information divergence formula to calculate the temporal spectral fitting value be-tween pixels and rubber plantation endmember, finally, extracted the rubber plantation information from fitting value image with the favorable threshold.Experimental results showed that the temporal spectral feature matching method could effectively extract rubber plantation information from the medium-resolution remote sensing image, and the accuracy was better than 93%, further-more, the spectral information divergence matching could achieve higher accuracy than that of the spectral angle.%纹理与物候是橡胶林区别于天然林及其他地类的有效诊断特征。对多时序中分辨率遥感影像合成的多谱影像提取出橡胶林端元光谱，通过光谱角及光谱信息散度计算像元时序光谱与橡胶林端元时序光谱间的匹配度，最后通过匹配度阈值的调整提取出橡胶林信息。实验结果表明，采用时序光谱特征匹配法能够有效地提取出中分辨率遥感影像中的橡胶林信息，精度在93％以上，且光谱信息散度较光谱角匹配能取得更高的精度。