光谱匹配与空间特征提取协同高光谱目标识别方法

如有对接意向，请电话垂询010-68918340

光谱匹配与空间特征提取协同高光谱目标识别方法

本发明涉及一种光谱匹配与空间特征提取协同高光谱目标识别方法，解决了单纯依赖基于光谱空间的识别方法不足问题。本方法首先利用直接法对高光谱影像进行几何校正，消除几何变形，然后利用利用经验线性法反演模型进行光谱重建，把成像光谱图像的辐射值DN转换为地物的反射率值，再选择最终成分光谱或参考光谱，并按照光谱角度匹配(SAM)规则，利用选择的最终成分光谱进行目标的初步识别分类，然后通过最佳指数模型选择最优波段组合，获取信息量最为丰富的多光谱图像。最后在光谱匹配基础上利用空间结构特征提取辅助目标精确识别，直至获得高光谱影像的更为精确结果。本发明可有效提高目标分类识别的精度，定量分类精度达到93％。

Spectralmatching and spatial feature extraction synergetic hyperspectral targetrecognition method

Theinvention relates to a spectral matching and spatial feature extractionsynergetic hyperspectral target recognition method which overcomes the defectsof simply relying on a recognition method based on a spectral space. The spectralmatching and spatial feature extraction synergetic hyperspectral targetrecognition method comprises the steps that a hyperspectral image is subjectedto geometric correction by utilizing a direct method firstly, geometric deformationis eliminated, then spectrum reconstruction is conducted by utilizing an empiricallinear method inversion model, a radiation value DN of an imaging spectralimage is converted into a reflectivity value of a ground feature, then a finalcomponent spectrum or a reference spectrum is selected, target primary recognitionand classification are conducted by utilizing the selected final componentspectrum according to a spectral angle mapping (SAM) rule, then an optimal wavebandcombination is selected through an optimal index model, and a multispectralimage with the most abundant information amount is obtained; and finally, auxiliarytarget precise recognition is extracted by utilizing the spatial structurefeature on the basis of spectral mapping until the more precise result of the hyperspectralimage is obtained. According to the spectral matching and spatial featureextraction synergetic hyperspectral target recognition method, the precision oftarget classification recognition can be effectively improved, and thequantitative classification precision reaches 93%.