【摘要】 鉴于在时频局部化能力方面小波包变换优于小波变换,将高光谱影像像元光谱曲线作为1维信号并对其进行多尺度小波包变换分解,得到不同尺度上的低频和高频成分向量。根据不同地物像元光谱小波包分解最佳基有很大差异,而同一地物像元光谱小波包分解的前若干个最佳基完全相同的特点,提出一种基于前若干个最佳小波包基特征参量数组的分类特征参量和目标识别方法,并对AVIRIS影像中的特征如地物植被、水体、岩石及某些阴影等进行提取与制图。更多还原

【Abstract】 The wavelet packet transformation has better time-frequency localization ability compared with the wavelet transformation. The pixel spectral curves of hyperspectral images as one dimension signals were decomposed by multi-scale wavelet packet transformations, and acquired different-scale component vectors in low and high frequency signals. According to the fact that the pixel spectra of different objects have different bases of best wavelet packet but the pixel spectra of same features are identical ones in the first some best wavelet packet groups, and by means of the wavelet packet decomposition of the spectral features of some objects such as vegetation, water, rock and shadow, a new target identification method has been probed for mapping features based on the composition of the characteristic parameters of the first several best wavelet packet bases for each pixel spectrum. The experiment results show that the features to vegetation, water, rock and shadow on AVIRIS image can be identified and mapped.更多还原