论文部分内容阅读
采用1维离散小波HAAR、DB4、SYM4对LOPEX 93数据库中的6条水稻反射光谱曲线进行10层小波分解。利用小波近似系数重构信号,采用步长行走法计算重构信号的小波分形维数。研究各尺度下小波分形维数、小波细节系数方差、小波细节系数信息熵、小波近似系数重构方差的特征。结果表明水稻光谱曲线具有分形特征,分形计算中相关系数值均大于0.9证明分形计算的有效性。4个参数的尺度特征揭示了水稻光谱曲线特征尺度转折点出现在尺度6,当水稻光谱分辨率小于64 nm,才能较好地反映光谱曲线峰谷细节特性。通过田间实测18条水稻光谱,计算各尺度的两种植被指数及植被指数与叶绿素的相关系数,进一步证明这一结论。
One-dimensional discrete wavelet HAAR, DB4 and SYM4 were used to decompose six rice reflectance spectra in LOPEX 93 database by 10-layer wavelet decomposition. The signal is reconstructed by using the wavelet approximation coefficient, and the wavelet fractal dimension of the reconstructed signal is calculated by the step walking method. The fractal dimension of wavelet, the variance of wavelet detail coefficient, the entropy of wavelet detail coefficient and the variance of wavelet approximation coefficient are studied. The results showed that the spectral curve of rice had fractal characteristics, and the correlation coefficients in fractal calculation were all greater than 0.9 to prove the validity of fractal calculation. The scale features of the four parameters revealed that the turning point of the characteristic curve of rice spectral curve appeared at the scale of 6, and the spectral peak-valley detail characteristics of rice spectral curve could be well reflected when the spectral resolution of rice was less than 64 nm. This conclusion is further confirmed by field measurements of 18 rice cultivars, calculation of two vegetation indices at various scales, and correlation coefficients between vegetation index and chlorophyll.