基于SAM遥感影像的分类技术研究

Research on classification Techniques of remote sensing image by SAM

目的以混合像元提纯为突破口,研究遥感影像高分类精度的方法,以减少由于地物波谱的复杂性、传感器空间分辨率的局限性导致混合像元普遍存在而引起的影像信息不确定性和分类精度低指标性。方法利用光谱角(SAM)分类法,经过混合像元辨别和提纯、端元样区的定义等步骤,对影像进行分类。结果与传统最大似然分类法比较实验结果表明,SAM方法对地表覆盖比较复杂的零散地区遥感影像分类具有较高的精度。结论因选择合适的方法消除了混合像元,SAM是一种有效的遥感影像智能分类方法,对提高影像分类精度具有极为重要的实践意义。

Aim Starting with purification of mixed pixels, to study the more higher classification precision technique, which is used to reduce the uncertainty and low precision of classification as a result of mixed pixels widely contained in the remote sensing image because of the complexity of the spectrum taken from the field and the limit spatial resolution of the sensor. Methods Using the Spectral Angle Mapping （SAM） classification algorithm, the image is classified by a series of processes, such as distinguishing between mixed pixels and pure pixels, purification of mixed pixels, definition of endmember sample site. Results The experiment result in comparison with traditional supervised classification algorithm proves that SAM has higher precision of classification for remote sensing image, especially in the regions containing multi type scattered land cover. Conclusion For the sake of a practicable method that can eliminate the mixed pixel, SAM is one of the effective methods in intelligent classification of remote sensing image, which has important pragmatic significance for improving the precision of classification.